This Is A Chapter of The Stacks Project, Version 4287507, Compiled On Oct 20, 2014

HOMOLOGICAL ALGEBRA
Contents
1. Introduction
2. Basic notions
3. Preadditive and additive categories
4. Karoubian categories
5. Abelian categories
6. Extensions
7. Additive functors
8. Localization
9. Serre subcategories
10. K-groups
11. Cohomological delta-functors
12. Complexes
13. Truncation of complexes
14. Homotopy and the shift functor
15. Graded objects
16. Filtrations
17. Spectral sequences
18. Spectral sequences: exact couples
19. Spectral sequences: differential objects
20. Spectral sequences: filtered differential objects
21. Spectral sequences: filtered complexes
22. Spectral sequences: double complexes
23. Injectives
24. Projectives
25. Injectives and adjoint functors
26. Essentially constant systems
27. Inverse systems
28. Exactness of products
29. Other chapters
References
1
2
2
4
5
12
13
16
18
20
22
24
28
30
33
34
40
41
43
45
47
50
55
55
56
58
59
63
63
64
1. Introduction
Basic homological algebra will be explained in this document. We add as needed
in the other parts, since there is clearly an infinite amount of this stuff around. A
reference is [ML63].
This is a chapter of the Stacks Project, version 4287507, compiled on Oct 20, 2014.
1
HOMOLOGICAL ALGEBRA
2. Basic notions
The following notions are considered basic and will not be defined, and or proved.
This does not mean they are all necessarily easy or well known.
(1) Nothing yet.
3. Preadditive and additive categories
Here is the definition of a preadditive category.
Definition 3.1. A category A is called preadditive if each morphism set MorA (x, y)
is endowed with the structure of an abelian group such that the compositions
Mor(x, y) Mor(y, z) Mor(x, z)
are bilinear. A functor F : A B of preadditive categories is called additive if and
only if F : Mor(x, y) Mor(F (x), F (y)) is a homomorphism of abelian groups for
all x, y Ob(A).
In particular for every x, y there exists at least one morphism x y, namely the
zero map.
Lemma 3.2. Let A be a preadditive category. Let x be an object of A. The
following are equivalent
(1) x is an initial object,
(2) x is a final object, and
(3) idx = 0 in MorA (x, x).
Furthermore, if such an object 0 exists, then a morphism : x y factors through
0 if and only if = 0.
Proof. Omitted.
Definition 3.3. In a preadditive category A we call zero object, and we denote it

0 any final and initial object as in Lemma 3.2 above.
Lemma 3.4. Let A be a preadditive category.
Let x, y Ob(A).`If the product
`
x y exists, then so does the coproduct x y.
If
` the coproduct x y exists, then
so does the product x y. In this case also x y
= x y.
Proof. Suppose that z = x y with projections p : z x and q : z y. Denote
i : x z the morphism corresponding to (1, 0). Denote j : y z the morphism
corresponding to (0, 1). Thus we have the commutative diagram
/x
?
?z
1
/ y
where the diagonal compositions are zero. It follows that i p + j q : z z is

the identity since it is a morphism which upon composing with p gives p and upon
composing with q gives q. Suppose given morphisms a : x w and b : y w.
HOMOLOGICAL ALGEBRA
Then we can form the map a p + b q : z w. In this `

way we get a bijection
Mor(z, w) = Mor(x, w) Mor(y, w) which show that z = x y.
`
We leave it to the reader to construct the morphisms p, q given a coproduct x y
instead of a product.

Definition 3.5. Given a pair of objects x, y in a preadditive category A we call
direct sum, and we denote it x y the product x y endowed with the morphisms
i, j, p, q as in Lemma 3.4 above.
Remark 3.6. Note that the proof of Lemma 3.4 shows that given p and q the
morphisms i, j are uniquely determined by the rules p i = idx , q j = idy ,
p j = 0, q i = 0. Moreover, we automatically have i p + j q = idxy . Similarly,
given i, j the morphisms p and q are uniquely determined. Finally, given objects
x, y, z and morphisms i : x z, j : y z, p : z x and q : z y such that
p i = idx , q j = idy , p j = 0, q i = 0 and i p + j q = idz , then z is the direct
sum of x and y with the four morphisms equal to i, j, p, q.
Lemma 3.7. Let A, B be preadditive categories. Let F : A B be an additive
functor. Then F transforms direct sums to direct sums and zero to zero.
Proof. Suppose F is additive. A direct sum z of x and y is characterized by having
morphisms i : x z, j : y z, p : z x and q : z y such that p i = idx ,
q j = idy , p j = 0, q i = 0 and i p + j q = idz , according to Remark 3.6.
Clearly F (x), F (y), F (z) and the morphisms F (i), F (j), F (p), F (q) satisfy exactly
the same relations (by additivity) and we see that F (z) is a direct sum of F (x) and
F (y).

Definition 3.8. A category A is called additive if it is preadditive and finite products exist, in other words it has a zero object and direct sums.
Namely the empty product is a finite product and if it exists, then it is a final
object.
Definition 3.9. Let A be a preadditive category. Let f : x y be a morphism.
(1) A kernel of f is a morphism i : z x such that (a) f i = 0 and (b) for any
i0 : z 0 x such that f i0 = 0 there exists a unique morphism g : z 0 z
such that i0 = i g.
(2) If the kernel of f exists, then we denote this Ker(f ) x.
(3) A cokernel of f is a morphism p : y z such that (a) p f = 0 and (b)
for any p0 : y z 0 such that p0 f = 0 there exists a unique morphism
g : z z 0 such that p0 = g p.
(4) If a cokernel of f exists we denote this y Coker(f ).
(5) If a kernel of f exists, then a coimage of f is a cokernel for the morphism
Ker(f ) x.
(6) If a kernel and coimage exist then we denote this x Coim(f ).
(7) If a cokernel of f exists, then the image of f is a kernel of the morphism
y Coker(f ).
(8) If a cokernel and image of f exist then we denote this Im(f ) y.
We first relate the direct sum to kernels as follows.
HOMOLOGICAL ALGEBRA
Lemma 3.10. Let C be a preadditive category. Let xy with morphisms i, j, p, q as

in Lemma 3.4 be a direct sum in C. Then i : x x y is a kernel of q : x y y.
Dually, p is a cokernel for j.
Proof. Let f : z x y be a morphism such that q f = 0. We have to show
that there exists a unique morphism g : z x such that f = i g. Since i p + j q
is the identity on x y we see that
f = (i p + j q) f = i p f
and hence g = p f works. Uniquess holds because p i is the identity on x. The
proof of the second statement is dual.

Lemma 3.11. Let f : x y be a morphism in a preadditive category such that the
kernel, cokernel, image and coimage all exist. Then f can be factored uniquely as
x Coim(f ) Im(f ) y.
Proof. There is a canonical morphism Coim(f ) y because Ker(f ) x y
is zero. The composition Coim(f ) y Coker(f ) is zero, because it is the
unique morphism which gives rise to the morphism x y Coker(f ) which is
zero. Hence Coim(f ) y factors uniquely through Im(f ) y, which gives us the
desired map.

Example 3.12. Let k be a field. Consider the category of filtered vector spaces
over k. (See Definition 16.1.) Consider the filtered vector spaces (V, F ) and (W, F )
with V = W = k and

V if i < 0
W if i 0
F iV =
and F i W =
0 if i 0
0 if i > 0
The map f : V W corresponding to idk on the underlying vector spaces has
trivial kernel and cokernel but is not an isomorphism. Note also that Coim(f ) = V
and Im(f ) = W . This means that the category of filtered vector spaces over k is
not abelian.
4. Karoubian categories
Skip this section on a first reading.
Definition 4.1. Let C be a preadditive category. We say C is Karoubian if every
idempotent endomorphism of an object of C has a kernel.
The dual notion would be that every idempotent endomorphism of an object has a
cokernel. However, in view of the (dual of the) following lemma that would be an
equivalent notion.
Lemma 4.2. Let C be a preadditive category. The following are equivalent
(1) C is Karoubian,
(2) every idempotent endomorphism of an object of C has a cokernel, and
(3) given an idempotent endomorphism p : z z of C there exists a direct sum
decomposition z = x y such that p corresponds to the projection onto y.
Proof. Assume (1) and let p : z z be as in (3). Let x = Ker(p) and y =
Ker(1 p). There are maps x z and y z. Since (1 p)p = 0 we see that
p : z z factors through y, hence we obtain a morphism z y. Similarly we obtain
a morphism z x. We omit the verification that these four morphisms induce an
HOMOLOGICAL ALGEBRA
isomorphism x = y z as in Remark 3.6. Thus (1) (3). The implication (2)

(3) is dual. Finally, condition (3) implies (1) and (2) by Lemma 3.10.

Lemma 4.3. Let D be a preadditive category.
(1) If D has countable products and kernels of maps which have a right inverse,
then D is Karoubian.
(2) If D has countable coproducts and cokernels of maps which have a left inverse, then D is Karoubian.
Proof. Let X be an object of D and let e : X X be an idempotent. The functor
e
W 7 Ker(MorD (W, X)
MorD (W, X))
if representable if and only if e has a kernel. Note that for any abelian group A
and idempotent endomorphism e : A A we have
Y
Y
Ker(e : A A) = Ker( :
A
A)
nN
nN
where
(a1 , a2 , a3 , . . .) = (ea1 + (1 e)a2 , ea2 + (1 e)a3 , . . .)
Moreover, has the right inverse
(a1 , a2 , a3 , . . .) = (a1 , (1 e)a1 + ea2 , (1 e)a2 + ea3 , . . .).
Hence (1) holds. The proof of (2) is dual (using the dual definition of a Karoubian
category, namely condition (2) of Lemma 4.2).

5. Abelian categories
An abelian category is a category satisfying just enough axioms so the snake lemma
holds. An axiom (that is sometimes forgotten) is that the canonical map Coim(f )
Im(f ) of Lemma 3.11 is always an isomorphism. Example 3.12 shows that it is
necessary.
Definition 5.1. A category A is abelian if it is additive, if all kernels and cokernels
exist, and if the natural map Coim(f ) Im(f ) is an isomorphism for all morphisms
f of A.
Lemma 5.2. Let A be a preadditive category. The additions on sets of morphisms
make Aopp into a preadditive category. Furthermore, A is additive if and only if
Aopp is additive, and A is abelian if and only if Aopp is abelian.
Proof. Omitted.
Definition 5.3. Let f : x y be a morphism in an abelian category.

(1) We say f is injective if Ker(f ) = 0.
(2) We say f is surjective if Coker(f ) = 0.
If x y is injective, then we say that x is a subobject of y and we use the notation
x y. If x y is surjective, then we say that y is a quotient of x.
Lemma 5.4. Let f : x y be a morphism in an abelian category. Then
(1) f is injective if and only if f is a monomorphism, and
(2) f is surjective if and only if f is an epimorphism.
Proof. Omitted.
HOMOLOGICAL ALGEBRA
In an abelian category, if x y is a subobject, then we denote

x/y = Coker(x y).
Lemma 5.5. Let A be an abelian category. All finite limits and finite colimits exist
in A.
Proof. To show that finite limits exist it suffices to show that finite products and
equalizers exist, see Categories, Lemma 18.4. Finite products exist by definition
and the equalizer of a, b : x y is the kernel of a b. The argument for finite
colimits is similar but dual to this.

Example 5.6. Let A be an abelian category. Pushouts and fibre products in A
have the following simple descriptions:
(1) If a : x y, b : z y are morphisms in A, then we have the fibre product:
x y z = Ker((a, b) : x z y).
(2) If a : y x, b : y z are morphisms in A, then we have the pushout:
x qy z = Coker((a, b) : y x z).
Definition 5.7. Let A be an additive category. We say a sequence of morphisms
... x y z ...
in A is a complex if the composition of any two (drawn) arrows is zero. If A is
abelian then we say a sequence as above is exact at y if Im(x y) = Ker(y z).
We say it is exact if it is exact at every object. A short exact sequence is an exact
complex of the form
0 A B C 0.
In the following lemma we assume the reader knows what it means for a sequence
of abelian groups to be exact.
Lemma 5.8. Let A be an abelian category. Let 0 M1 M2 M3 0 be a
complex of A.
(1) M1 M2 M3 0 is exact if and only if
0 HomA (M3 , N ) HomA (M2 , N ) HomA (M1 , N )
is an exact sequence of abelian groups for all objects N of A, and
(2) 0 M1 M2 M3 is exact if and only if
0 HomA (N, M1 ) HomA (N, M2 ) HomA (N, M1 )
is an exact sequence of abelian groups for all objects N of A.
Proof. Omitted. Hint: See Algebra, Lemma 10.1.
Definition 5.9. Let A be an abelian category. Let i : A B and q : B C be

morphisms of A such that 0 A B C 0 is a short exact sequence. We say
the short exact sequence is split if there exist morphisms j : C B and p : B A
such that (B, i, j, p, q) is the direct sum of A and C.
Lemma 5.10. Let A be an abelian category. Let 0 A B C 0 be a short
exact sequence.
(1) Given a morphism s : C B left inverse to B C, there exists a unique
: B A such that (s, ) splits the short exact sequence as in Definition
5.9.
HOMOLOGICAL ALGEBRA
(2) Given a morphism : B A right inverse to A B, there exists a unique

s : C B such that (s, ) splits the short exact sequence as in Definition
5.9.
Proof. Omitted.
Lemma 5.11. Let A be an abelian category. Let

w
/y

x
/ z
be a commutative diagram.
(1) The diagram is cartesian if and only if
(g,f )
(k,h)
0 w x y z
is exact.
(2) The diagram is cocartesian if and only if
(g,f )
(k,h)
w x y z 0
is exact.
Proof. Let u = (g, f ) : w x y and v = (k, h) : x y z. Let p : x y x
and q : x y y be the canonical projections. Let i : Ker(v) x y be the
canonical injection. By Example 5.6, the diagram is cartesian if and only if there
exists an isomorphism r : Ker(v) w with f r = q i and g r = p i. The
u
v
sequence 0 w x y z is exact if and only if there exists an isomorphism
r : Ker(v) w with u r = i. But given r : Ker(v) w, we have f r = q i and
g r = p i if and only if q u r = f r = q i and p u r = g r = p i, hence
if and only if u r = i. This proves (1), and then (2) follows by duality.

w
/y

x
/ z
(1) If the diagram is cartesian, then the morphism Ker(f ) Ker(k) induced
by g is an isomorphism.
(2) If the diagram is cocartesian, then the morphism Coker(f ) Coker(k)
induced by h is an isomorphism.
Proof. Suppose the diagram is cartesian. Let e : Ker(f ) Ker(k) be induced by
g. Let i : Ker(f ) w and j : Ker(k) x be the canonical injections. There exists
t : Ker(k) w with f t = 0 and g t = j. Hence, there exists u : Ker(k) Ker(f )
with iu = t. It follows giue = gte = je = gi and f iue = 0 = f i, hence
i u e = i. Since i is a monomorphism this implies u e = idKer(f ) . Furthermore,
we have j e u = g i u = g t = j. Since j is a monomorphism this implies
e u = idKer(k) . This proves (1). Now, (2) follows by duality.
HOMOLOGICAL ALGEBRA

w
/y

x
/ z
(1) If the diagram is cartesian and k is an epimorphism, then the diagram is
cocartesian and f is an epimorphism.
(2) If the diagram is cocartesian and g is a monomorphism, then the diagram
is cartesian and h is a monomorphism.
Proof. Suppose the diagram is cartesian and k is an epimorphism. Let u = (g, f ) :
w x y and let v = (k, h) : x y z. As k is an epimorphism, v is an
u
v
epimorphism, too. Therefore and by Lemma 5.11, the sequence 0 w x y
z 0 is exact. Thus, the diagram is cocartesian by Lemma 5.11. Finally, f is an
epimorphism by Lemma 5.12 and Lemma 5.4. This proves (1), and (2) follows by
duality.

Lemma 5.14. Let A be an abelian category.
(1) If x y is surjective, then for every z y the projection x y z z is
surjective.
(2) If x y is injective, then for every x z the morphism z z qx y is
injective.
Proof. Immediately from Lemma 5.4 and Lemma 5.13.
Lemma 5.15. Let A be an abelian category. Let f : x y and g : y z be

morphisms with g f = 0. Then, the following statements are equivalent:
f
(1) The sequence x y z is exact.

(2) For every h : w y with g h = 0 there exist an object v, an epimorphism
k : v w and a morphism l : v x with h k = f l.
Proof. Let i : Ker(g) y be the canonical injection. Let p : x Coim(f ) be the
canonical projection. Let j : Im(f ) Ker(g) be the canonical injection.
Suppose (1) holds. Let h : w y with g h = 0. There exists c : w Ker(g) with
i c = h. Let v = x Ker(g) w with canonical projections k : v w and l : v x,
so that c k = p l. Then, h k = i c k = i j p l = f l. As j p is an
epimorphism by hypothesis, k is an epimorphism by Lemma 5.13. This implies (2).
Suppose (2) holds. Then, g i = 0. So, there are an object w, an epimorphism
k : w Ker(g) and a morphism l : w x with f l = i k. It follows i j p l =
f l = i k. Since i is a monomorphism we see that j p l = k is an epimorphism.
So, j is an epimorphisms and thus an isomorphism. This implies (1).

f
/y

u
/ v
/z
/ w
HOMOLOGICAL ALGEBRA
(1) If the first row is exact and k is a monomorphism, then the induced sequence
Ker() Ker() Ker() is exact.
(2) If the second row is exact and g is an epimorphism, then the induced sequence Coker() Coker() Coker() is exact.
Proof. Suppose the first row is exact and k is a monomorphism. Let a : Ker()
Ker() and b : Ker() Ker() be the induced morphisms. Let h : Ker()
x, i : Ker() y and j : Ker() z be the canonical injections. As j is a
monomorphism we have b a = 0. Let c : s Ker() with b c = 0. Then,
gic = j bc = 0. By Lemma 5.15 there are an object t, an epimorphism d : t s
and a morphism e : t x with icd = f e. Then, ke = f e = icd = 0.
As k is a monomorphism we get e = 0. So, there exists m : t Ker() with
hm = e. It follows iam = f hm = f e = icd. As i is a monomorphism we
get am = cd. Thus, Lemma 5.15 implies (1), and then (2) follows by duality.
/y
/ u
/ v
/0
/z
/ w
be a commutative diagram with exact rows.

(1) There exists a unique morphism : Ker() Coker() such that the
diagram
yo

v
/ 0
y z Ker()
/ Ker()
Coker() qu v o

Coker()
commutes, where and 0 are the canonical projections and and 0 are
the canonical coprojections.
(2) The induced sequence
f0
g0
k0
l0
Ker() Ker() Ker() Coker() Coker() Coker()

is exact. If f is injective then so is f 0 , and if l is surjective then so is l0 .
Proof. As is an epimorphism and is a monomorphism by Lemma 5.13, uniqueness of is clear. Let p = y z Ker() and q = Coker() qu v. Let h : Ker() y,
i : Ker() z and j : Ker() p be the canonical injections. Let p : u Coker()
be the canonical projection. Keeping in mind Lemma 5.13 we get a commutative
10
HOMOLOGICAL ALGEBRA
diagram with exact rows

/ Ker()
/p
/ Ker()
/ z
/0
/ v
/ w

/ Coker()
/ u
/ y
/0
/ q
As l 0 = i = 0 and as the third row of the diagram above is exact,

there is an a : p u with k a = 0 . As the upper right quadrangle of the
diagram above is cartesian, Lemma 5.12 yields an epimorphism b : x Ker()
with 0 j b = f . It follows k a j b = 0 j b = f = k . As k
is a monomorphism this implies a j b = . It follows p a j b = p = 0.
As b is an epimorphism this implies p a j = 0. Therefore, as the top row of the
diagram above is exact, there exists : Ker() Coker() with = p a. It
follows = p a = 0 k a = 0 0 as desired.
As the upper right quadrangle in the diagram above is cartesian there is a c :
Ker() p with 0 c = h and c = g 0 . It follows g 0 = c =
0 0 c = 0 h = 0. As is a monomorphism this implies g 0 = 0.
Next, let d : r Ker() with d = 0. Applying Lemma 5.15 to the exact
sequence p Ker() 0 and d yields an object s, an epimorphism m : s r and

a morphism n : s p with n = d m. As p a n = d m = 0, applying
p
Lemma 5.15 to the exact sequence x u Coker() and a n yields an object

t, an epimorphism : t s and a morphism : t x with a n = . It
holds 0 n = k = f . Let = 0 n f : t y.
Then, = 0. It follows that there is a : t Ker() with = h . It holds
i g 0 = g h = g 0 n g f = i n = i d m . As i
is a monomorphism we get g 0 = d m . Thus, as m is an epimorphism,
g0
Lemma 5.15 implies that Ker() Ker() Coker() is exact. Then, the claim
follows by Lemma 5.16 and duality.

/y
x
x0

/ u0
/ u
/ y0
0

/ v0
/ v
/z
/ z0
~
/ w0
/0
/ w
/0
HOMOLOGICAL ALGEBRA
11
be a commutative diagram with exact rows. Then, the induced diagram

Ker()
/ Ker()
/ Ker()
/ Coker()
/ Coker()
/ Coker()

Ker(0 )

/ Ker( 0 )

/ Ker( 0 )

/ Coker(0 )

/ Coker( 0 )

/ Coker( 0 )
commutes.
Proof. Omitted.

/x
/y
w

/ x0

w0
/z

/ y0

/ z0
be a commutative diagram with exact rows.

(1) If , are surjective and is injective, then is surjective.
(2) If , are injective and is surjective, then is injective.
Proof. Assume , are surjective and is injective. We may replace w0 by
Im(w0 x0 ), i.e., we may assume that w0 x0 is injective. We may replace z
by Im(y z), i.e., we may assume that y z is surjective. Then we may apply
Lemma 5.17 to
/z
/0
/y
Ker(y z)

/ Ker(y 0 z 0 )

/ y0

/ z0
to conclude that Ker(y z) Ker(y 0 z 0 ) is surjective. Finally, we apply

Lemma 5.17 to
/ Ker(y z)
/0
/x
w

/ w0

/ x0

/ Ker(y 0 z 0 )
to conclude that x x0 is surjective. This proves (1). The proof of (2) is dual to
this.

/x
/y
/w
v

v0

/ w0

/ x0
/z

/ y0

/ z0
be a commutative diagram with exact rows. If , are isomorphisms, is injective,

and is surjective then is an isomorphism.
Proof. Immediate consequence of Lemma 5.19.
12
HOMOLOGICAL ALGEBRA
6. Extensions
Definition 6.1. Let A be an abelian category. Let A, C Ob(A). An extension
E of B by A is a short exact sequence
0 A E B 0.
By abuse of language we often omit mention of the morphisms A E and E B,
although they are definitively part of the structure of an extension.
Definition 6.2. Let A be an abelian category. Let A, B Ob(A). The set of
isomorphism classes of extensions of B by A is denoted
ExtA (B, A).
This is called the Ext-group.
This definition works, because by our conventions A is a set, and hence ExtA (B, A)
is a set. In any of the cases of big abelian categories listed in Categories, Remark
2.2. one can check by hand that ExtA (B, A) is a set as well. Also, we will see
later that this is always the case when A has either enough projectives or enough
injectives. Insert future reference here.
Actually we can turn ExtA (, ) into a functor
Aopp A Sets,
(A, B) 7 ExtA (A, B)
as follows:
(1) Given a morphism B 0 B and an extension E of B by A we define
E 0 = E B B 0 so that we have the following commutative diagram of short
exact sequences
0
/A
/ E0
/ B0
/0

/A

/E

/B
/0
The extension E 0 is called the pullback of E via B 0 B.

0
0
(2) Given
` a morphism A A and an extension E of B by A we define E =
A0 A E so that we have the following commutative diagram of short exact
sequences
/E
/B
/0
/A
0

/ A0
/0
/ E0
/B
0
0
The extension E is called the pushout of E via A A0 .
To see that this defines a functor as indicated above there are several things to verify.
First of all functoriality in the variable B requires that (E B B 0 )B 0 B 00 = E B B 00
which is a general property of fibre products. Dually one deals with functoriality
in the variable A. Finally, given A A0 and B 0 B we have to show that
a
a
A0
(E B B 0 )
E) B B 0
= (A0
A
A
`
as extensions of B 0 by A0 . Recall that A0 A E is a quotient of A0 E. Thus the
right hand side is a quotient of A0 E B B 0 , and it is straightforward to see that
the kernel is exactly what you need in order to get the left hand side.
HOMOLOGICAL ALGEBRA
13
Note that if E1 and E2 are extensions of B by A, then E1 E2 is an extension of

B B by A A. We pull back by the diagonal map B B B and we push out
by the sum map A A A to get an extension E1 + E2 of B by A.
0
/ AA
/ E1 E2
/ BB
/0

/ E0
O

/ BB
O
/0

/A
O
/A
/ E1 + E2
/B
/0
The extension E1 + E2 is called the Baer sum of the given extensions.

Lemma 6.3. The construction (E1 , E2 ) 7 E1 + E2 above defines a commutative
group law on ExtA (B, A) which is functorial in both variables.
Proof. Omitted.
Lemma 6.4. Let A be an abelian category. Let 0 M1 M2 M3 0 be a

short exact sequence in A.
(1) There is a canonical six term exact sequence of abelian groups
0
/ HomA (M3 , N )
ExtA (M3 , N )
/ HomA (M2 , N )
/ HomA (M1 , N )
/ ExtA (M2 , N )
/ ExtA (M1 , N )
for all objects N of A, and

(2) there is a canonical six term exact sequence of abelian groups
0
/ HomA (N, M1 )
ExtA (N, M1 )
/ HomA (N, M2 )
/ HomA (N, M3 )
/ ExtA (N, M2 )
/ ExtA (N, M3 )
for all objects N of A.

Proof. Omitted. Hint: The boundary maps are defined using either the pushout
or pullback of the given short exact sequence.

7. Additive functors
Recall that we defined, in Categories, Definition 23.1 the notion of a right exact,
left exact and exact functor in the setting of a functor between categories that
have finite (co)limits. Thus this applies in particular to functors between abelian
categories.
Lemma 7.1. Let A and B be abelian categories. Let F : A B be a functor.
(1) If F is either left or right exact, then it is additive.
(2) If F is additive then it is left exact if and only if for every short exact
sequence 0 A B C 0 the sequence 0 F (A) F (B) F (C)
is exact.
14
HOMOLOGICAL ALGEBRA
(3) If F is additive then it is right exact if and only if for every short exact
sequence 0 A B C 0 the sequence F (A) F (B) F (C) 0
is exact.
(4) If F is additive then it is exact if and only if for every short exact sequence
0 A B C 0 the sequence 0 F (A) F (B) F (C) 0 is
exact.
Proof. Let us first note that if F commutes with the empty limit or the empty
colimit, then F (0) = 0. In particular F applied to the zero morphism is zero. We
will use this below without mention.
Suppose that F is left exact, i.e., commutes with finite limits. Then F (A A) =
F (A) F (A) with projections F (p) and F (q). Hence F (A A) = F (A) F (A)
with all four morphisms F (i), F (j), F (p), F (q) equal to their counterparts in B as
they satisfy the same relations, see Remark 3.6. Then f = F (p + q) is a morphism
f : F (A) F (A) F (A) such that f F (i) = F (p i + q i) = F (idA ) = idF (A) .
And similarly f F (j) = idA . We conclude that F (p + q) = F (p) + F (q). For any
pair of morphisms a, b : B A the map g = F (i a + j b) : F (B) F (A) F (A)
is a morphism such that F (p) g = F (p (i a + j b)) = F (a) and similarly
F (q) g = F (b). Hence g = F (i) F (a) + F (j) F (b). The sum of a and b is the
composition
B
ia+jb
/ AA
p+q
/ A.
Applying F we get
F (B)
F (i)F (a)+F (j)F (b)
/ F (A) F (A)
F (p)+F (q)
/ A.
where we used the expressions for f and g obtained above. Hence F is additive.1
Denote f : B C a map from B to C. Exactness of 0 A B C just means
that A = Ker(f ). Clearly the kernel of f is the equalizer of the two maps f and
0 from B to C. Hence if F commutes with limits, then F (Ker(f )) = Ker(F (f ))
which exactly means that 0 F (A) F (B) F (C) is exact.
Conversely, suppose that F is additive and transforms any short exact sequence
0 A B C into an exact sequence 0 F (A) F (B) F (C). Because it
is additive it commutes with direct sums and hence finite products in A. To show
it commutes with finite limits it therefore suffices to show that it commutes with
equalizers. But equalizers in an abelian category are the same as the kernel of the
difference map, hence it suffices to show that F commutes with taking kernels. Let
f : A B be a morphism. Factor f as A I B with f 0 : A I surjective
and i : I B injective. (This is possible by the definition of an abelian category.)
Then it is clear that Ker(f ) = Ker(f 0 ). Also 0 Ker(f 0 ) A I 0 and
0 I B B/I 0 are short exact. By the condition imposed on F we see
that 0 F (Ker(f 0 )) F (A) F (I) and 0 F (I) F (B) F (B/I) are exact.
Hence it is also the case that F (Ker(f 0 )) is the kernel of the map F (A) F (B),
and we win.
The proof of (3) is similar to the proof of (2). Statement (4) is a combination of
(2) and (3).

1Im sure there is an infinitely slicker proof of this.
HOMOLOGICAL ALGEBRA
15
Lemma 7.2. Let A and B be abelian categories. Let F : A B be an exact

functor. For every pair of objects A, B of A the functor F induces an abelian group
homomorphism
ExtA (B, A) ExtB (F (B), F (A))
which maps the extension E to F (E).
Proof. Omitted.
The following lemma is used in the proof that the category of abelian sheaves on a
site is abelian, where the functor b is sheafification.
Lemma 7.3. Let a : A B and b : B A be functors. Assume that
(1) A, B are additive categories, a, b are additive functors, and a is right adjoint
to b,
(2) B is abelian and b is left exact, and
(3) ba
= idA .
Then A is abelian.
Proof. As B is abelian we see that all finite limits and colimits exist in B by Lemma
5.5. Since b is a left adjoint we see that b is also right exact and hence exact, see
Categories, Lemma 24.5. Let : B1 B2 be a morphism of B. In particular, if
K = Ker(B1 B2 ), then K is the equalizer of 0 and and hence bK is the equalizer
of 0 and b, hence bK is the kernel of b. Similarly, if Q = Coker(B1 B2 ), then
Q is the coequalizer of 0 and and hence bQ is the coequalizer of 0 and b, hence
bQ is the cokernel of b. Thus we see that every morphism of the form b in A has
a kernel and a cokernel. However, since ba
= id we see that every morphism of A
is of this form, and we conclude that kernels and cokernels exist in A. In fact, the
argument shows that if : A1 A2 is a morphism then
Ker() = bKer(a),
and
Coker() = bCoker(a).
Now we still have to show that Coim() = Im(). We do this as follows. First
note that since A has kernels and cokernels it has all finite limits and colimits (see
proof of Lemma 5.5). Hence we see by Categories, Lemma 24.5 that a is left exact
and hence transforms kernels (=equalizers) into kernels.
Coim() = Coker(Ker() A1 )
by definition
= bCoker(a(Ker() A1 ))
by formula above
= bCoker(Ker(a) aA1 ))
a preserves kernels
= bCoim(a)
= bIm(a)
= bKer(aA2 Coker(a))
= Ker(baA2 bCoker(a))
= Ker(A2 bCoker(a))
= Ker(A2 Coker())
= Im()
Thus the lemma holds.
by definition
B is abelian
by definition
b preserves kernels
ba = idA
by formula above
by definition
16
HOMOLOGICAL ALGEBRA
8. Localization
In this section we note how Gabriel-Zisman localization interacts with the additive
structure on a category.
Lemma 8.1. Let C be a preadditive category. Let S be a left or right multiplicative system. There exists a canonical preadditive structure on S 1 C such that the
localization functor Q : C S 1 C is additive.
Proof. We will prove this in the case S is a left multiplicative system. The case
where S is a right multiplicative system is dual. Suppose that X, Y are objects of
C and that , : X Y are morphisms in S 1 C. According to Categories, Lemma
25.3 we may represent these by pairs s1 f, s1 g with common denominator s. In
this case we define + to be the equivalence class of s1 (f + g). In the rest of
the proof we show that this is well defined and that composition is bilinear. Once
this is done it is clear that Q is an additive functor.
Let us show construction above is well defined. An abstract way of saying this is
that filtered colimits of abelian groups agree with filtered colimits of sets and to use
Categories, Equation (25.5.1). We can work this out in a bit more detail as follows.
Say s : Y Y1 and f, g : X Y1 . Suppose we have a second representation of
, as (s0 )1 f 0 , (s0 )1 g 0 with s0 : Y Y2 and f 0 , g 0 : X Y2 . By Categories,
Remark 25.5 we can find a morphism s3 : Y Y3 and morphisms a1 : Y1 Y3 ,
a2 : Y2 Y3 such that a1 s = s3 = a2 s0 and also a1 f = a2 f 0 and a1 g = a2 g 0 .
Hence we see that s1 (f + g) is equivalent to
1
s1
3 (a1 (f + g)) = s3 (a1 f + a1 g)
0
0
= s1
3 (a2 f + a2 g )
0
0
= s1
3 (a2 (f + g ))
which is equivalent to (s0 )1 (f 0 + g 0 ).

Fix s : Y Y 0 and f, g : X Y 0 with = s1 f and = s1 g as morphisms
X Y in S 1 C. To show that composition is bilinear first consider the case of a
morphism : Y Z in S 1 C. Say = t1 h for some h : Y Z 0 and t : Z Z 0
in S. Using LMS2 we choose morphisms a : Y 0 Z 00 and t0 : Z 0 Z 00 in S such
that a s = t0 h. Picture
Z
t

/ Z0

/ Z 00
t0
f,g

/ Y0
Then = (t0 t)1 (a f ) and = (t0 t)1 (a g). Hence we see that
( + ) is represented by (t0 t)1 (a (f + g)) = (t0 t)1 (a f + a g) which
represents + .
Finally, assume that : W X is another morphism of S 1 C. Say = r1 i for
some i : W X 0 and r : X X 0 in S. We claim that we can find a morphism
HOMOLOGICAL ALGEBRA
17
s : Y 0 Y 00 in S and morphisms a00 , b00 : X 0 Y 00 such that the following diagram

commutes
Y
s
X
s
f,g,f +g

/ Y0
s0
a00 ,b00 ,a00 +b00

i /
/ Y 00
W
X0
Namely, using LMS2 we can first choose s1 : Y 0 Y1 , s2 : Y 0 Y2 in S and
a : X 0 Y1 , b : X 0 Y2 such that a s = s1 f and b s = s2 f . Then
using that the category Y 0 /S is filtered (see Categories, Remark 25.5), we can find
a s0 : Y 0 Y 00 and morphisms a0 : Y1 Y 00 , b0 : Y2 Y 00 such that s0 = a0 s1
and s0 = b0 s2 . Setting a00 = a0 a and b00 = b0 b works. At this point we see that
the compositions and are represented by (s0 s)1 a00 and (s0 s)1 b00 .
Hence + is represented by (s0 s)1 (a00 + b00 ) which by the diagram again
is a representative of ( + ) .

Lemma 8.2. Let C be an additive category. Let S be a left or right multiplicative
system. Then S 1 C is an additive category and the localization functor Q : C
S 1 C is additive.
Proof. By Lemma 8.1 we see that S 1 C is preadditive and that Q is additive.
Recall that the functor Q commutes with finite colimits (resp. finite limits), see
Categories, Lemmas 25.7 and 25.14. We conclude that S 1 C has a zero object and
direct sums, see Lemmas 3.2 and 3.4.

The following lemma describes the kernel (see Definition 9.5) of the localization
functor in case we invert a multiplicative system.
Lemma 8.3. Let C be an additive category. Let S be a multiplicative system. Let
X be an object of C. The following are equivalent
(1) Q(X) = 0 in S 1 C,
(2) there exists Y Ob(C) such that 0 : X Y is an element of S, and
(3) there exists Z Ob(C) such that 0 : Z X is an element of S.
Proof. If (2) holds we see that 0 = Q(0) : Q(X) Q(Y ) is an isomorphism. In
the additive category S 1 C this implies that Q(X) = 0. Hence (2) (1). Similarly,
(3) (1). Suppose that Q(X) = 0. This implies that the morphism f : 0 X
is transformed into an isomorphism in S 1 C. Hence by Categories, Lemma 25.18
there exists a morphism g : Z 0 such that f g S. This proves (1) (3).
Similarly, (1) (2).

(1) If S is a left multiplicative system, then the category S 1 A has cokernels
and the functor Q : A S 1 A commutes with them.
(2) If S is a right multiplicative system, then the category S 1 A has kernels
and the functor Q : A S 1 A commutes with them.
(3) If S is a multiplicative system, then the category S 1 A is abelian and the
functor Q : A S 1 A is exact.
18
HOMOLOGICAL ALGEBRA
Proof. Assume S is a left multiplicative system. Let a : X Y be a morphism

of S 1 A. Then a = s1 f for some s : Y Y 0 in S and f : X Y 0 . Since Q(s)
is an isomorphism we see that the existence of Coker(a : X Y ) is equivalent to
the existence of Coker(Q(f ) : X Y 0 ). Since Coker(Q(f )) is the coequalizer of 0
and Q(f ) we see that Coker(Q(f )) is represented by Q(Coker(f )) by Categories,
Lemma 25.7. This proves (1).
Part (2) is dual to part (1).
If S is a multiplicative system, then S is both a left and a right multiplicative
system. Thus we see that S 1 A has kernels and cokernels and Q commutes with
kernels and cokernels. To finish the proof of (3) we have to show that Coim = Im
in S 1 A. Again using that any arrow in S 1 A is isomorphic to an arrow Q(f ) we
see that the result follows from the result for A.

9. Serre subcategories
In [Ser53, Chapter I, Section 1] a notion of a class of abelian groups is defined.
This notion has been extended to abelian categories by many authors (in slightly
different ways). We will use the following variant which is virtually identical to
Serres original definition.
Definition 9.1. Let A be an abelian category.
(1) A Serre subcategory of A is a nonempty full subcategory C of A such that
given an exact sequence
ABC
with A, C Ob(C), then also B Ob(C).
(2) A weak Serre subcategory of A is a nonempty full subcategory C of A such
that given an exact sequence
A0 A1 A2 A3 A4
with A0 , A1 , A3 , A4 in C, then also A2 in C.
In some references the second notion is called a thick subcategory and in other
references the first notion is called a thick subcategory. However, it seems that
the notion of a Serre subcategory is universally accepted to be the one defined
above. Note that in both cases the category C is abelian and that the inclusion
functor C A is a fully faithful exact functor. Lets characterize these types of
subcategories in more detail.
Lemma 9.2. Let A be an abelian category. Let C be a subcategory of A. Then C
is a Serre subcategory if and only if the following conditions are satisfied:
(1)
(2)
(3)
(4)
0 Ob(C),
C is a strictly full subcategory of A,
any subobject or quotient of an object of C is an object of C,
if A Ob(A) is an extension of objects of C then also A Ob(C).
Moreover, a Serre subcategory is an abelian category and the inclusion functor is

exact.
Proof. Omitted.
HOMOLOGICAL ALGEBRA
19
Lemma 9.3. Let A be an abelian category. Let C be a subcategory of A. Then C

is a weak Serre subcategory if and only if the following conditions are satisfied:
(1) 0 Ob(C),
(2) C is a strictly full subcategory of A,
(3) kernels and cokernels in A of morphisms between objects of C are in C,
(4) if A Ob(A) is an extension of objects of C then also A Ob(C).
Moreover, a weak Serre subcategory is an abelian category and the inclusion functor
is exact.
Proof. Omitted.
Lemma 9.4. Let A, B be abelian categories. Let F : A B be an exact functor.

Then the full subcategory of objects C of A such that F (C) = 0 forms a Serre
subcategory of A.
Proof. Omitted.
Definition 9.5. Let A, B be abelian categories. Let F : A B be an exact

functor. Then the full subcategory of objects C of A such that F (C) = 0 is called
the kernel of the functor F , and is sometimes denoted Ker(F ).
Lemma 9.6. Let A be an abelian category. Let C A be a Serre subcategory.
There exists an abelian category A/C and an exact functor
F : A A/C
which is essentially surjective and whose kernel is C. The category A/C and the
functor F are characterized by the following universal property: For any exact
functor G : A B such that C Ker(G) there exists a factorization G = H F
for a unique exact functor H : A/C B.
Proof. Consider the set of arrows of A defined by the following formula
S = {f Arrows(A) | Ker(f ), Coker(f ) Ob(C)}.
We claim that S is a multiplicative system. To prove this we have to check MS1,
MS2, MS3, see Categories, Definition 25.1.
It is clear that identities are elements of S. Suppose that f : A B and g : B C
are elements of S. There are exact sequences
0 Ker(f ) Ker(gf ) Ker(g)
Coker(f ) Coker(gf ) Coker(g) 0
Hence it follows that gf S. This proves MS1. (In fact, a similar argument will
show that S is a saturated multiplicative system, see Categories, Definition 25.17.)
Consider a solid diagram
A

C
/B

/ C qA B
with t S. Set W = C qA B = Coker((t, g) : A C B). Then Ker(t) Ker(s)

is surjective and Coker(t) Coker(s) is an isomorphism. Hence s is an element of
S. This proves LMS2 and the proof of RMS2 is dual.
20
HOMOLOGICAL ALGEBRA
Finally, consider morphisms f, g : B C and a morphism s : A B in S

such that f s = g s. This means that (f g) s = 0. In turn this means
that I = Im(f g) C is a quotient of Coker(s) hence an object of C. Thus
t : C C 0 = C/I is an element of S such that t (f g) = 0, i.e., such that
t f = t g. This proves LMS3 and the proof of RMS3 is dual.
Having proved that S is a multiplicative system we set A/C = S 1 A, and we set
F equal to the localization functor Q. By Lemma 8.4 the category A/C is abelian
and F is exact. If X is in the kernel of F = Q, then by Lemma 8.3 we see that
0 : X Z is an element of S and hence X is an object of C, i.e., the kernel of F
is C. Finally, if G is as in the statement of the lemma, then G turns every element
of S into an isomorphism. Hence we obtain the functor H : A/C B from the
universal property of localization, see Categories, Lemma 25.6.

Lemma 9.7. Let A, B be abelian categories. Let F : A B be an exact functor.
Let C = Ker(F ). Then the induced functor F : A/C B is faithful.
Proof. This is true because the kernel of F is zero by construction. Namely, if
f : X Y is a morphism in A/C such that F (f ) = 0, then Ker(f ) X and
Y Coker(f ) are transformed into isomorphisms by F , hence are isomorphisms
by the remark on the kernel of F . Thus f = 0.

10. K-groups
Definition 10.1. Let A be an abelian category. We denote K0 (A) the zeroth Kgroup of A. It is the abelian group constructed as follows. Take the free abelian
group on the objects on A and for every short exact sequence 0 A B C 0
impose the relation [B] [A] [C] = 0.
Another way to say this is that there is a presentation
M
M
Z[A B C]
Z[A] K0 (A) 0
ABC ses
AOb(A)
with [A B C] 7 [B] [A] [C] of K0 (A). The short exact sequence

0 0 0 0 0 leads to the relation [0] = 0 in K0 (A). There are no settheoretical issues as all of our categories are small if not mentioned otherwise.
Some examples of K-groups for categories of modules over rings where computed
in Algebra, Section 53.
Lemma 10.2. Let F : A B be an exact functor between abelian categories.
Then F induces a homomorphism of K-groups K0 (F ) : K0 (A) K0 (B) by simply
setting K0 (F )([A]) = [F (A)].
Proof. Proves itself.
Suppose we are given an object M of an abelian category A and a complex of the

form
(10.2.1)
...
/M
/M
/M
/M
/ ...
In this situation we define

H 0 (M, , ) = Ker()/Im(),
and H 1 (M, , ) = Ker()/Im().
HOMOLOGICAL ALGEBRA
21
Lemma 10.3. Let A be an abelian category. Let C A be a Serre subcategory and

set B = A/C.
(1) The exact functors C A and A B induce an exact sequence
K0 (C) K0 (A) K0 (B) 0
of K-groups, and
(2) the kernel of K0 (C) K0 (A) is equal to the collection of elements of the
form
[H 0 (M, , )] [H 1 (M, , )]
where (M, , ) is a complex as in (10.2.1) with the property that it becomes
exact in B; in other words that H 0 (M, , ) and H 1 (M, , ) are objects
of C.
Proof. We omit the proof of (1). The proof of (2) is in a sense completely combinatorial. First we remark that any class of the type [H 0 (M, , )] [H 1 (M, , )]
is zero in K0 (A) by the following calculation
0 = [M ] [M ]
= [Ker()] + [Im()] [Ker()] [Im()]
= [Ker()/Im()] [Ker()/Im()]
= [H 1 (M, , )] [H 0 (M, , )]
as desired. Hence it suffices to show that any element in the kernel of K0 (C)
K0 (A) is of this form.
Any element x in K0 (C) can be represented as the difference x = [P ] [Q] of two
objects of C (fun exercise). Suppose that this element maps to zero in K0 (A). This
means that there exist
`
(1) a finite set I = I + I ,
(2) for each i I a short exact sequence
0 Ai B i C i 0
in the abelian category A
such that
[P ] [Q] =
X
iI +
([Bi ] [Ai ] [Ci ])
X
iI
([Bi ] [Ai ] [Ci ])
in the free abelian group on the objects of A. We can rewrite this as

X
X
X
X
[P ] +
([Ai ] + [Ci ]) +
[Bi ] = [Q] +
([Ai ] + [Ci ]) +
+
iI
iI
iI
iI +
[Bi ].
Since the right and left hand side should contain the same objects of A counted
with multiplicity, this means there should be a bijection between the terms which
occur above. Set
a
a
T + = {p}
{a, c} I +
{b} I
and
T = {q}
{a, c} I
{b} I + .
22
Set T = T +
HOMOLOGICAL ALGEBRA
T = {p, q}
`
{a, b, c} I.
Q
O(t) = Ai
Bi
Ci
For t T define
if
t=p
if
t=q
if t = (a, i)
if t = (b, i)
if t = (c, i)
Hence we can view : T + T as a bijection such that O(t) = O( (t)) for all
+
+
t T + . Let t
be the unique element such that (t+
0 = (p) and let t0 T
0 ) = q.
Consider the object
M
O(t)
M+ =
+
tT
By using we see that it is equal to the object

M
M =
O(t)
tT
Consider the map

: M + M
which on the summand O(t) = Ai corresponding to t = (a, i), i I + uses the
map Ai Bi into the summand O((b, i)) = Bi of M and on the summand
O(t) = Bi corresponding to (b, i), i I uses the map Bi Ci into the summand
O((c, i)) = Ci of M . The map is zero on the summands corresponding to p and
(c, i), i I + . Similarly, consider the map
: M M +
which on the summand O(t) = Ai corresponding to t = (a, i), i I uses the
map Ai Bi into the summand O((b, i)) = Bi of M + and on the summand
O(t) = Bi corresponding to (b, i), i I + uses the map Bi Ci into the summand
O((c, i)) = Ci of M + . The map is zero on the summands corresponding to q and
(c, i), i I .
Note that the kernel of is equal to the direct sum of the summand P and the
summands O((c, i)) = Ci , i I + and the subobjects Ai inside the summands
O((b, i)) = Bi , i I . The image of is equal to the direct sum of the summands
O((c, i)) = Ci , i I + and the subobjects Ai inside the summands O((b, i)) = Bi ,
i I . In other words we see that
P
= Ker()/Im().
In exactly the same way we see that
Q
= Ker()/Im().
Since as we remarked above the existence of the bijection shows that M + = M
we see that the lemma follows.

11. Cohomological delta-functors
Definition 11.1. Let A, B be abelian categories. A cohomological -functor or
simply a -functor from A to B is given by the following data:
(1) a collection F n : A B, n 0 of additive functors, and
(2) for every short exact sequence 0 A B C 0 of A a collection
ABC : F n (C) F n+1 (A), n 0 of morphisms of B.
These data are assumed to satisfy the following axioms
HOMOLOGICAL ALGEBRA
23
(1) for every short exact sequence as above the sequence

0
/ F 0 (A)
/ F 0 (B)
u
ABC
/ F 1 (B)
F (A)
F (A)
/ F 0 (C)
/ F 1 (C)
ABC
/ F 2 (B)
/ ...
is exact, and
(2) for every morphism (A B C) (A0 B 0 C 0 ) of short exact
sequences of A the diagrams
F n (C)

F n (C 0 )
ABC
A0 B 0 C 0
/ F n+1 (A)

/ F n+1 (A0 )
are commutative.
Note that this in particular implies that F 0 is left exact.
Definition 11.2. Let A, B be abelian categories. Let (F n , F ) and (Gn , G ) be
-functors from A to B. A morphism of -functors from F to G is a collection of
transformation of functors tn : F n Gn , n 0 such that for every short exact
sequence 0 A B C 0 of A the diagrams
F n (C)
F,ABC
tn

Gn (C)
G,ABC
/ F n+1 (A)

tn+1
/ Gn+1 (A)
are commutative.
Definition 11.3. Let A, B be abelian categories. Let F = (F n , F ) be a -functor
from A to B. We say F is a universal -functor if an only if for every -functor
G = (Gn , G ) and any morphism of functors t : F 0 G0 there exists a unique
morphism of -functors {tn }n0 : F G such that t = t0 .
Lemma 11.4. Let A, B be abelian categories. Let F = (F n , F ) be a -functor from
A to B. Suppose that for every n > 0 and any A Ob(A) there exists an injective
morphism u : A B (depending on A and n) such that F n (u) : F n (A) F n (B)
is zero. Then F is a universal -functor.
Proof. Let G = (Gn , G ) be a -functor from A to B and let t : F 0 G0 be
a morphism of functors. We have to show there exists a unique morphism of functors {tn }n0 : F G such that t = t0 . We construct tn by induction on n.
For n = 0 we set t0 = t. Suppose we have already constructed a unique sequence of
transformation of functors ti for i n compatible with the maps in degrees n.
Let A Ob(A). By assumption we may choose a embedding u : A B such
that F n+1 (u) = 0. Let C = B/u(A). The long exact cohomology sequence for
24
HOMOLOGICAL ALGEBRA
the short exact sequence 0 A B C 0 and the -functor F gives that

F n+1 (A) = Coker(F n (B) F n (C)) by our choice of u. Since we have already
defined tn we can set
tn+1
: F n+1 (A) Gn+1 (A)
A
equal to the unique map such that
Coker(F n (B) F n (C))
F,ABC
/ Coker(Gn (B) Gn (C))
tn
tn+1
A
F n+1 (A)
G,ABC
/ Gn+1 (A)
commutes. This is clearly uniquely determined by the requirements imposed. We

omit the verification that this defines a transformation of functors.

Lemma 11.5. Let A, B be abelian categories. Let F : A B be a functor. If
there exists a universal -functor (F n , F ) from A to B with F 0 = F , then it is
determined up to unique isomorphism of -functors.
Proof. Immediate from the definitions.
12. Complexes
Of course the notions of a chain complex and a cochain complex are dual and you
only have to read one of the two parts of this section. So pick the one you like.
(Actually, this doesnt quite work right since the conventions on numbering things
are not adapted to an easy transition between chain and cochain complexes.)
A chain complex A in an additive category A is a complex
dn+1
n
An1 . . .
. . . An+1 An
of A. In other words, we are given an object Ai of A for all i Z and for all i Z
a morphism di : Ai Ai1 such that di1 di = 0 for all i. A morphism of chain
complexes f : A B is given by a family of morphisms fi : Ai Bi such that
all the diagrams
/ Ai1
Ai
di
fi

Bi
di
fi1
/ Bi1
commute. The category of chain complexes of A is denoted Ch(A). The full subcategory consisting of objects of the form
. . . A2 A1 A0 0 0 . . .
is denoted Ch0 (A). In other words, a chain complex A belongs to Ch0 (A) if
and only if Ai = 0 for all i < 0. A homotopy h between a pair of morphisms of
chain complexes f, g : A B is is a collection of morphisms hi : Ai Bi+1 such
that we have
fi gi = di+1 hi + hi1 di
for all i. Clearly, the notions of chain complex, morphism of chain complexes,
and homotopies between morphisms of chain complexes makes sense even in a
preadditive category.
HOMOLOGICAL ALGEBRA
25
Lemma 12.1. Let A be an additive category. Let f, g : B C be morphisms

of chain complexes. Suppose given morphisms of chain complexes a : A B ,
and c : C D . If {hi : Bi Ci+1 } defines a homotopy between f and g, then
{ci+1 hi ai } defines a homotopy between c f a and c g a.
Proof. Omitted.
In particular this means that it makes sense to define the category of chain complexes with maps up to homotopy. Well return to this later.
Definition 12.2. Let A be an additive category. We say a morphism a : A B
is a homotopy equivalence if there exists a morphism b : B A such that there
exists a homotopy between a b and idA and there exists a homotopy between b a
and idB . If there exists such a morphism between A and B , then we say that A
and B are homotopy equivalent.
In other words, two complexes are homotopy equivalent if they become isomorphic
in the category of complexes up to homotopy.
(1) The category of chain complexes in A is abelian.
(2) A morphism of complexes f : A B is injective if and only if each
fn : An Bn is injective.
(3) A morphism of complexes f : A B is surjective if and only if each
fn : An Bn is surjective.
(4) A sequence of chain complexes
f
A
B
C
is exact at B if and only if each sequence
fi
gi
Ai Bi Ci
is exact at Bi .
Proof. Omitted.
For any i Z the ith homology group of a chain complex A in an abelian category
is defined by the following formula
Hi (A ) = Ker(di )/Im(di+1 ).
If f : A B is a morphism of chain complexes of A then we get an induced
morphism Hi (f ) : Hi (A ) Hi (B ) because clearly fi (Ker(di : Ai Ai1 ))
Ker(di : Bi Bi1 ), and similarly for Im(di+1 ). Thus we obtain a functor
Hi : Ch(A) A.
(1) A morphism of chain complexes f : A B is called a quasi-isomorphism
if the induced maps Hi (f ) : Hi (A ) Hi (B ) is an isomorphism for all
i Z.
(2) A chain complex A is called acyclic if all of its homology objects Hi (A )
are zero.
26
HOMOLOGICAL ALGEBRA
(1) If the maps f, g : A B are homotopic, then the induced maps Hi (f )

and Hi (g) are equal.
(2) If the map f : A B is a homotopy equivalence, then f is a quasiisomorphism.
Proof. Omitted.
Lemma 12.6. Let A be an abelian category. Suppose that

0 A B C 0
is a short exact sequence of chain complexes of A. Then there is a canonical long
exact homology sequence
...
Hi (A )
Hi1 (A )
...
...
/ Hi (B )
/ Hi (C )
/ Hi1 (B )
/ Hi1 (C )
...
...
... s
Proof. Omitted. The maps come from the Snake Lemma 5.17 applied to the
diagrams
/ Bi /Im(dB,i+1 )
Ai /Im(dA,i+1 )

dA,i
/ Ker(dA,i1 )
/ Ci /Im(dC,i+1 )
dB,i
/ Ker(dB,i1 )
/0
dC,i
/ Ker(dC,i1 )
A cochain complex A in an additive category A is a complex

dn1
dn
. . . An1 An An+1 . . .
of A. In other words, we are given an object Ai of A for all i Z and for all i Z
a morphism di : Ai Ai+1 such that di+1 di = 0 for all i. A morphism of cochain
complexes f : A B is given by a family of morphisms f i : Ai B i such that
all the diagrams
/ Ai+1
Ai
di
f i+1

di /
Bi
B i+1
commute. The category of cochain complexes of A is denoted CoCh(A). The full
subcategory consisting of objects of the form
. . . 0 0 A0 A1 A2 . . .
is denoted CoCh0 (A). In other words, a cochain complex A belongs to the
subcategory CoCh0 (A) if and only if Ai = 0 for all i < 0. A homotopy h between
HOMOLOGICAL ALGEBRA
27
a pair of morphisms of cochain complexes f, g : A B is is a collection of

morphisms hi : Ai B i1 such that we have
f i g i = di1 hi + hi+1 di
for all i. Clearly, the notions of cochain complex, morphism of cochain complexes,
and homotopies between morphisms of cochain complexes makes sense even in a
preadditive category.
Lemma 12.7. Let A be an additive category. Let f, g : B C be morphisms of
cochain complexes. Suppose given morphisms of cochain complexes a : A B ,
and c : C D . If {hi : B i C i1 } defines a homotopy between f and g, then
{ci1 hi ai } defines a homotopy between c f a and c g a.
Proof. Omitted.
In particular this means that it makes sense to define the category of cochain
complexes with maps up to homotopy. Well return to this later.
Definition 12.8. Let A be an additive category. We say a morphism a : A B
is a homotopy equivalence if there exists a morphism b : B A such that there
exists a homotopy between a b and idA and there exists a homotopy between b a
and idB . If there exists such a morphism between A and B , then we say that A
and B are homotopy equivalent.
In other words, two complexes are homotopy equivalent if they become isomorphic
in the category of complexes up to homotopy.
(1) The category of cochain complexes in A is abelian.
(2) A morphism of cochain complexes f : A B is injective if and only if
each f n : An B n is injective.
(3) A morphism of cochain complexes f : A B is surjective if and only if
each f n : An B n is surjective.
(4) A sequence of cochain complexes
f
A
B
C
is exact at B if and only if each sequence
fi
gi
Ai B i C i
is exact at B i .
Proof. Omitted.
For any i Z the ith cohomology group of a cochain complex A is defined by the
following formula
H i (A ) = Ker(di )/Im(di1 ).
If f : A B is a morphism of cochain complexes of A then we get an induced
morphism H i (f ) : H i (A ) H i (B ) because clearly f i (Ker(di : Ai Ai+1 ))
Ker(di : B i B i+1 ), and similarly for Im(di1 ). Thus we obtain a functor
H i : CoCh(A) A.
28
HOMOLOGICAL ALGEBRA
(1) A morphism of cochain complexes f : A B of A is called a quasiisomorphism if the induced maps H i (f ) : H i (A ) H i (B ) is an isomorphism for all i Z.
(2) A cochain complex A is called acyclic if all of its cohomology objects
H i (A ) are zero.
(1) If the maps f, g : A B are homotopic, then the induced maps H i (f )
and H i (g) are equal.
(2) If f : A B is a homotopy equivalence, then f is a quasi-isomorphism.
Proof. Omitted.
Lemma 12.12. Let A be an abelian category. Suppose that

0 A B C 0
is a short exact sequence of chain complexes of A. Then there is a canonical long
exact homology sequence
...
H i (A )
H i+1 (A )
...
...
/ H i (B )
/ H i (C )
/ H i+1 (B )
/ H i+1 (C )
...
...
... s
Proof. Omitted. The maps come from the Snake Lemma 5.17 applied to the
diagrams
Ai /Im(di1
A )
diA

/ Ker(di+1 )
A
/ B i /Im(di1 )
B
/ C i /Im(di1 )
C
diB

/ Ker(di+1 )
B
/0
diC

/ Ker(di+1 )
C
13. Truncation of complexes

Let A be an abelian category. Let A be a chain complex. There are several ways
to truncate the complex A .
(1) The stupid truncation n is the subcomplex n A defined by the rule
(n A )i = 0 if i > n and (n A )i = Ai if i n. In a picture
n A
...
/0
/ An
/ An1
/ ...

A
...

/ An+1

/ An

/ An1
/ ...
Note the property n A /n1 A = An [n].
HOMOLOGICAL ALGEBRA
29
(2) The stupid truncation n is the quotient complex n A defined by the

rule (n A )i = Ai if i n and (n A )i = 0 if i < n. In a picture
A
...
/ An+1
/ An
/ An1
/ ...

n A
...

/ An+1

/ An

/0
/ ...
The map of complexes n A n+1 A is surjective with kernel An [n].

(3) The canonical truncation n A is defined by the picture
n A
...
/ An+1
/ Ker(dn )
/0
/ ...

A
...

/ An+1

/ An

/ An1
/ ...
Note that these complexes have the property that

Hi (A ) if i n
Hi (n A ) =
0
if i < n
A
...
/ An+1
/ An
/ An1
/ ...

n A
...

/0

/ Coker(dn+1 )

/ An1
/ ...

Hi (A ) if i n
Hi (n A ) =
0
if i > n
Let A be an abelian category. Let A be a cochain complex. There are four ways
to truncate the complex A .
(1) The stupid truncation n is the subcomplex n A defined by the rule
(n A )i = 0 if i < n and (n A )i = Ai if i n. In a picture
n A
...
/0
/ An
/ An+1
/ ...

A
...

/ An1

/ An

/ An+1
/ ...
Note the property n A /n+1 A = An [n].

(2) The stupid truncation n is the quotient complex n A defined by the
rule (n A )i = 0 if i > n and (n A )i = Ai if i n. In a picture
A
...
/ An1
/ An
/ An+1
/ ...

n A
...

/ An1

/ An

/0
/ ...
The map of complexes n A n1 A is surjective with kernel An [n].
30
HOMOLOGICAL ALGEBRA

n A
...
/ An1
/ Ker(dn )
/0
/ ...

A
...

/ An1

/ An

/ An+1
/ ...

i
H (A ) if i n
i
H (n A ) =
0
if i > n
A
...
/ An1
/ An
/ An+1
/ ...

n A
...

/0

/ Coker(dn1 )

/ An+1
/ ...

0
if i < n
i
H (n A ) =
H i (A ) if i n
14. Homotopy and the shift functor
It is an annoying feature that signs and indices have to be part of any discussion
of homological algebra2.
Definition 14.1. Let A be an additive category. Let A be a chain complex with
boundary maps dA,n : An An1 . For any k Z we define the k-shifted chain
complex A[k] as follows:
(1) we set A[k]n = An+k , and
(2) we set dA[k],n : A[k]n A[k]n1 equal to dA[k],n = (1)k dA,n+k .
If f : A B is a morphism of chain complexes, then we let f [k] : A[k] B[k]
be the morphism of chain complexes with f [k]n = fk+n .
Of course this means we have functors [k] : Ch(A) Ch(A) which mutually
commute (on the nose, without any intervening isomorphisms of functors), such
that A[k][l] = A[k + l] and with [0] = idCh(A) .
Definition 14.2. Let A be an abelian category. Let A be a chain complex with
boundary maps dA,n : An An1 . For any k Z we identify Hi+k (A )
Hi (A[k] ) via the identification Ai+k = A[k]i .
This identification is functorial in A . Note that since no signs are involved in
this definition we actually get a compatible system of identifications of all the homology objects Hik (A[k] ), which are further compatible with the identifications
A[k][l] = A[k + l] and with [0] = idCh(A) .
Let A be an additive category. Suppose that A and B are chain complexes,
a, b : A B are morphisms of chain complexes, and {hi : Ai Bi+1 } is a
homotopy between a and b. Recall that this means that ai bi = di+1 hi +hi1 di .
2I am sure you think that my conventions are wrong. If so and if you feel strongly about it
then drop me an email with an explanation.
HOMOLOGICAL ALGEBRA
31
What if a = b? Then we obtain the formula 0 = di+1 hi + hi1 di , in other words,

di+1 hi = hi1 di . By definition above this means the collection {hi } above
defines a morphism of chain complexes
A B[1] .
Such a thing is the same as a morphism A[1] B by our remarks above. This
proves the following lemma.
Lemma 14.3. Let A be an additive category. Suppose that A and B are chain
complexes. Given any morphism of chain complexes a : A B there is a bijection
between the set of homotopies from a to a and MorCh(A) (A , B[1] ). More generally,
the set of homotopies between a and b is either empty or a principal homogeneous
space under the group MorCh(A) (A , B[1] ).
Proof. See above.

0 A B C 0
be a sort exact sequence of complexes. Suppose that {sn : Cn Bn } is a family
of morphisms which split the short exact sequences 0 An Bn Cn 0. Let
n : Bn An be the associated projections, see Lemma 5.10. Then the family of
morphisms
n1 dB,n sn : Cn An1
define a morphism of complexes (s) : C A[1] .
Proof. Denote i : A B and q : B C the maps of complexes in the
short exact sequence. Then in1 n1 dB,n sn = dB,n sn sn1 dC,n .
Hence in2 dA,n1 n1 dB,n sn = dB,n1 (dB,n sn sn1 dC,n ) =
dB,n1 sn1 dC,n as desired.

Lemma 14.5. Notation and assumptions as in Lemma 14.4 above. The morphism
of complexes (s) : C A[1] induces the maps
Hi ((s)) : Hi (C ) Hi (A[1] ) = Hi1 (A )
which occur in the long exact homology sequence associated to the short exact sequence of chain complexes by Lemma 12.6.
Proof. Omitted.
Lemma 14.6. Notation and assumptions as in Lemma 14.4 above. Suppose {s0n :
Cn Bn } is a second choice of splittings. Write s0n = sn + in hn for some
unique morphisms hn : Cn An . The family of maps {hn : Cn A[1]n+1 } is a
homotopy between the associated morphisms (s), (s0 ) : C A[1] .
Proof. Omitted.
Definition 14.7. Let A be an additive category. Let A be a cochain complex

with boundary maps dnA : An An+1 . For any k Z we define the k-shifted
cochain complex A[k] as follows:
(1) we set A[k]n = An+k , and
(2) we set dnA[k] : A[k]n A[k]n+1 equal to dnA[k] = (1)k dn+k
A .
32
HOMOLOGICAL ALGEBRA
If f : A B is a morphism of cochain complexes, then we let f [k] : A[k] B[k]

be the morphism of cochain complexes with f [k]n = f k+n .
Of course this means we have functors [k] : CoCh(A) CoCh(A) which mutually
commute (on the nose, without any intervening isomorphisms of functors) and such
that A[k][l] = A[k + l] and with [0] = idCoCh(A) .
Definition 14.8. Let A be an abelian category. Let A be a cochain complex
with boundary maps dnA : An An+1 . For any k Z we identify H i+k (A )
H i (A[k] ) via the identification Ai+k = A[k]i .
This identification is functorial in A . Note that since no signs are involved in
this definition we actually get a compatible system of identifications of all the homology objects H ik (A[k] ), which are further compatible with the identifications
A[k][l] = A[k + l] and with [0] = idCoCh(A) .
Let A be an additive category. Suppose that A and B are cochain complexes,
a, b : A B are morphisms of cochain complexes, and {hi : Ai B i1 } is a
homotopy between a and b. Recall that this means that ai bi = di1 hi +hi+1 di .
What if a = b? Then we obtain the formula 0 = di1 hi + hi+1 di , in other words,
di1 hi = hi+1 di . By definition above this means the collection {hi } above
defines a morphism of cochain complexes
A B[1] .
Such a thing is the same as a morphism A[1] B by our remarks above. This
proves the following lemma.
Lemma 14.9. Let A be an additive category. Suppose that A and B are cochain
complexes. Given any morphism of cochain complexes a : A B there is a
bijection between the set of homotopies from a to a and MorCoCh(A) (A , B[1] ).
More generally, the set of homotopies between a and b is either empty or a principal
homogeneous space under the group MorCoCh(A) (A , B[1] ).
Proof. See above.
Lemma 14.10. Let A be an additive category. Let

0 A B C 0
be a complex (!) of complexes. Suppose that we are given splittings B n = An C n
compatible with the maps in the displayed sequence. Let sn : C n B n and n :
B n An be the corresponding maps. Then the family of morphisms
n+1 dnB sn : C n An+1
define a morphism of complexes : C A[1] .
Proof. Denote i : A B and q : B C the maps of complexes in the
short exact sequence. Then in+1 n+1 dnB sn = dnB sn sn+1 dnC . Hence
in+2 dn+1
n+1 dnB sn = dn+1
(dnB sn sn+1 dnC ) = dn+1
sn+1 dnC as
A
B
B
desired.

Lemma 14.11. Notation and assumptions as in Lemma 14.10 above. Assume in
addition that A is abelian. The morphism of complexes : C A[1] induces the
maps
H i () : H i (C ) H i (A[1] ) = H i+1 (A )
HOMOLOGICAL ALGEBRA
33
which occur in the long exact homology sequence associated to the short exact sequence of cochain complexes by Lemma 12.12.
Proof. Omitted.
Lemma 14.12. Notation and assumptions as in Lemma 14.10. Let : A B ,

: B C be the given morphisms of complexes. Suppose (s0 )n : C n B n and
( 0 )n : B n An is a second choice of splittings. Write (s0 )n = sn + n hn and
( 0 )n = n + g n n for some unique morphisms hn : C n An and g n : C n An .
Then
(1) g n = hn , and
(2) the family of maps {g n : C n A[1]n1 } is a homotopy between , 0 : C
n
A[1] , more precisely ( 0 )n = n + g n+1 dnC + dn1
A[1] g .
Proof. As (s0 )n and ( 0 )n are splittings we have ( 0 )n (s0 )n = 0. Hence
0 = ( n + g n n ) (sn + n hn ) = g n n sn + n n hn = g n + hn
which proves (1). We compute ( 0 )n as follows
( n+1 + g n+1 n+1 ) dnB (sn + n hn ) = n + g n+1 dnC + dnA hn
n
Since hn = g n and since dn1
A[1] = dA we conclude that (2) holds.
15. Graded objects

We make the following definition.
Definition 15.1. Let A be an additive category. The category of graded objects of
A, denoted Gr(A), is the category with
(1) objects A = (Ai ) are families of objects Ai , i Z of objects of A, and
(2) morphisms f : A = (Ai ) B = (B i ) are families of morphisms f i : Ai
B i of A.
If A has countable direct sums, then we can asssociated to an object A = (Ai ) of
Gr(A) the object
M
A=
Ai
iZ
and set k A = A . In this case Gr(A) is equivalent to the category of pairs (A, k)
consisting of an object A of A and a direct sum decomposition
M
A=
ki A
iZ
by direct summands indexed by Z and a morphism (A, k) (B, k) of such objects

is given by a morphism : A B of A such that (k i A) k i B for all i
Z. Whenever our additive category A has countable direct sums we will use this
equivalence without further mention.
However, with our definitions an additive or abelian category does not necessarily
have all (countable) direct sums. In this case our definition still makes sense. For
example, if A = Vectk is the category of finite dimensional vector spaces over a
field k, then Gr(Vectk ) is the category of vector spaces with a given gradation
all of whose graded pieces are finite dimensional, and not the category of finite
dimensional vector spaces with a given graduation.
34
HOMOLOGICAL ALGEBRA
Lemma 15.2. Let A be an abelian category. The category of graded objects Gr(A)
is abelian.
Proof. Let f : A = (Ai ) B = (B i ) be a morphism of graded objects of A
given by morphisms f i : Ai B i of A. Then we have Ker(f ) = (Ker(f i )) and
Coker(f ) = (Coker(f i )) in the category Gr(A). Since we have Im = Coim in A we
see the same thing holds in Gr(A).

Remark 15.3 (Warning). There are abelian categories A having countable direct
sums but where countable direct sums are not exact. An example is the opposite
of the category of abelian sheaves on R. Namely, the category of abelian sheaves
on R has countable products, but countable
are not exact. For such a
L products
category the functor Gr(A) A, (Ai ) 7
Ai described above is not exact. It
is still true that Gr(A) is equivalent to the category of graded objects (A, k) of A,
but the kernel in the category of graded objects of a map : (A, k) (B, k) is
not equal to Ker() endowed with a direct sum decomposition, but rather it is the
direct sum of the kernels of the maps k i A k i B.
Definition 15.4. Let A be an additive category. If A = (Ai ) is a graded object,
then the kth shift A[k] is the graded object with A[k]i = Ak+i .
If A and B are graded objects of A, then we have
(15.4.1)
HomGr(A) (A, B[k]) = HomGr(A) (A[k], B)
and an element of this group is sometimes called a map of graded objects homogeneous of degree k.
Given any set G we can define G-graded objects of A as the category whose objects
are A = (Ag )gG families of objects parametrized by elements of G. Morphisms
f : A B are defined as families of maps f g : Ag B g where g runs over the
elements of G. If G is an abelian group, then we can (unambiguously) define shift
functors [g] on the category of G-graded objects by the rule (A[g])g0 = Ag+g0 . A
particular case of this type of construction is when G = Z Z. In this case the
objects of the category are called bigraded objects of A. The (p, q) component of
a bigraded object A is usually denoted Ap,q . For (a, b) Z Z we write A[a, b] in
stead of A[(a, b)]. A morphism A A[a, b] is sometimes called a map of bidegree
(a, b).
16. Filtrations
A nice reference for this material is [Del71, Section 1]. (Note that our conventions
regarding abelian categories are different.)
(1) A decreasing filtration F on an object A is a family (F n A)nZ of subobjects
of A such that
A . . . F n A F n+1 A . . . 0
(2) A filtered object of A is pair (A, F ) consisting of an object A of A and a
decreasing filtration F on A.
(3) A morphism (A, F ) (B, F ) of filtered objects is given by a morphism
: A B of A such that (F i A) F i B for all i Z.
(4) The category of filtered objects is denoted Fil(A).
HOMOLOGICAL ALGEBRA
35
(5) Given a filtered object (A, F ) and a subobject X A the induced filtration
on X is the filtration with F n X = X F n A.
(6) Given a filtered object (A, F ) and a surjection : A Y the quotient
filtration is the filtration with F n Y = (F n A).
(7) A filtration F on an object A is said to be finite if there exist n, m such
that F n A = A and F m A = 0.
T
(8) Given a filtered object (A, F ) we say F i A exists
if there exists a biggest
S
subobject of A contained in all F i A. We say F i A exists if there exists a
smallest subobject of A containing all F i A.
T
(9) The filtration on aSfiltered object (A, F ) is said to be separated if i F i A = 0
and exhaustive if F i A = A.
By abuse of notation we say that a morphism f : (A, F ) (B, F ) of filtered objects
is injective if f : A B is injective in the abelian category A. Similarly we say
f is surjective if f : A B is surjective in the category A. Being injective (resp.
surjective) is equivalent to being a monomorphism (resp. epimorphism) in Fil(A).
By Lemma 16.2 this is also equivalent to having zero kernel (resp. cokernel).
Lemma 16.2. Let A be an abelian category. The category of filtered objects Fil(A)
has the following properties:
(1)
(2)
(3)
(4)
It is an additive category.
It has a zero object.
It has kernels and cokernels, images and coimages.
In general it is not an abelian category.
Proof. It is clear that Fil(A) is additive with direct sum given by (A, F )(B, F ) =
(AB, F ) where F p (AB) = F p AF p B. The kernel of a morphism f : (A, F )
(B, F ) of filtered objects is the injection Ker(f ) A where Ker(f ) is endowed with
the induced filtration. The cokernel of a morphism f : A B of filtered objects
is the surjection B Coker(f ) where Coker(f ) is endowed with the quotient
filtration. Since all kernels and cokernels exist, so do all coimages and images. See
Example 3.12 for the last statement.

Definition 16.3. Let A be an abelian category. A morphism f : A B of filtered
objects of A is said to be strict if f (F i A) = f (A) F i B for all i Z.
This also equivalent to requiring that f 1 (F i B) = F i A + Ker(f ) for all i Z. We
characterize strict morphisms as follows.
Lemma 16.4. Let A be an abelian category. Let f : A B be a morphism of
filtered objects of A. The following are equivalent
(1) f is strict,
(2) the morphism Coim(f ) Im(f ) of Lemma 3.11 is an isomorphism.
Proof. Note that Coim(f ) Im(f ) is an isomorphism of objects of A, and that
part (2) signifies that it is an isomorphism of filtered objects. By the description
of kernels and cokernels in the proof of Lemma 16.2 we see that the filtration
on Coim(f ) is the quotient filtration coming from A Coim(f ). Similarly, the
filtration on Im(f ) is the induced filtration coming from the injection Im(f )
B. The definition of strict is exactly that the quotient filtration is the induced
filtration.
36
HOMOLOGICAL ALGEBRA
Lemma 16.5. Let A be an abelian category. Let f : A B be a strict monomorphism of filtered objects. Let g : A C be a morphism of filtered objects. Then
f g : A B C is a strict monomorphism.
Proof. Clear from the definitions.
Lemma 16.6. Let A be an abelian category. Let f : B A be a strict epimorphism

of filtered objects. Let g : C A be a morphism of filtered objects. Then f g :
B C A is a strict epimorphism.
Proof. Clear from the definitions.
Lemma 16.7. Let A be an abelian category. Let (A, F ), (B, F ) be filtered objects.
Let u : A B be a morphism of filtered objects. If u is injective then u is strict if
and only if the filtration on A is the induced filtration. If u is surjective then u is
strict if and only if the filtration on B is the quotient filtration.
Proof. This is immediate from the definition.
Lemma 16.8. Let A be an abelian category. Let f : A B, g : B C be strict

morphisms of filtered objects.
(1) In general the composition g f is not strict.
(2) If g is injective, then g f is strict.
(3) If f is surjective, then g f is strict.
Proof. Let B a vector space over a field k with basis e1 , e2 , with the filtration
F n B = B for n < 0, with F 0 B = ke1 , and F n B = 0 for n > 0. Now take
A = k(e1 + e2 ) and C = B/ke2 with filtrations induced by B, i.e., such that A B
and B C are strict (Lemma 16.7). Then F n (A) = A for n < 0 and F n (A) = 0
for n 0. Also F n (C) = C for n 0 and F n (C) = 0 for n > 0. So the (nonzero)
composition A C is not strict.
Assume g is injective. Then
g(f (F p A)) = g(f (A) F p B)
= g(f (A)) g(F p (B))
= (g f )(A) (g(B) F p C)
= (g f )(A) F p C.
The first equality as f is strict, the second because g is injective, the third because
g is strict, and the fourth because (g f )(A) g(B).
Assume f is surjective. Then
(g f )1 (F i C) = f 1 (F i B + Ker(g))
= f 1 (F i B) + f 1 (Ker(g))
= F i A + Ker(f ) + Ker(g f )
= F i A + Ker(g f )
The first equality because g is strict, the second because f is surjective, the third
because f is strict, and the last because Ker(f ) Ker(g f ).

The following lemma says that subobjects of a filtered object have a well defined
filtration independent of a choice of writing the object as a cokernel.
HOMOLOGICAL ALGEBRA
37
Lemma 16.9. Let A be an abelian category. Let (A, F ) be a filtered object of A.

Let X Y A be subobjects of A. On the object
Y /X = Ker(A/X A/Y )
the quotient filtration coming from the induced filtration on Y and the induced
filtration coming from the quotient filtration on A/X agree. Any of the morphisms
X Y , X A, Y A, Y A/X, Y Y /X, Y /X A/X are strict (with
induced/quotient filtrations).
Proof. The quotient filtration Y /X is given by F p (Y /X) = F p Y /(X F p Y ) =
F p Y /F p X because F p Y = Y F p A and F p X = X F p A. The induced filtration
from the injection Y /X A/X is given by
F p (Y /X) = Y /X F p (A/X)
= Y /X (F p A + X)/X
= (Y F p A)/(X F p A)
= F p Y /F p X.
Hence the first statement of the lemma. The proof of the other cases is similar.
Lemma 16.10. Let A be an abelian category. Let A, B, C Fil(A). Let f : A B
and g : A C be morphisms Then there exists a pushout
A
/B
g0

C

/ C qA B
in Fil(A). If f is strict, so is f 0 .
Proof. Set C qA B equal to Coker((1, 1) : A C B) in Fil(A). This cokernel
exists, by Lemma 16.2. It is a pushout, see Example 5.6. Note that F p (C A B) is
the image of F p C F p B. Hence
(f 0 )1 (F p (C A B)) = g(f 1 (F p B))) + F p C
Whence the last statement.
Lemma 16.11. Let A be an abelian category. Let A, B, C Fil(A). Let f : B A

and g : C A be morphisms Then there exists a pushout
B A C
f0
g0

C
/B
g

/A
in Fil(A). If f is strict, so is f 0 .
Proof. This lemma is dual to Lemma 16.10.
Let A be an abelian category. Let (A, F ) be a filtered object of A. We denote

grpF (A) = grp (A) the object F p A/F p+1 A of A. This defines an additive functor
grp : Fil(A) A,
(A, F ) 7 grp (A).
38
HOMOLOGICAL ALGEBRA
Recall that we have defined the category Gr(A) of graded objects of A in Section
15. For (A, F ) in Fil(A) we may set
gr(A) = the graded object of A whose pth graded piece is grp (A)
and if A has countable direct sums, then we simply have
M
gr(A) =
grp (A)
This defines an additive functor
gr : Fil(A) Gr(A),
(A, F ) 7 gr(A).

(1) Let A be a filtered object and X A. Then for each p the sequence
0 grp (X) grp (A) grp (A/X) 0
is exact (with induced filtration on X and quotient filtration on A/X).
(2) Let f : A B be a morphism of filtered objects of A. Then for each p the
sequences
0 grp (Ker(f )) grp (A) grp (Coim(f )) 0
and
0 grp (Im(f )) grp (B) grp (Coker(f )) 0
are exact.
Proof. We have F p+1 X = X F p+1 A, hence map grp (X) grp (A) is injective.
Dually the map grp (A) grp (A/X) is surjective. The kernel of F p A/F p+1 A
A/X + F p+1 A is clearly F p+1 A + X F p A/F p+1 A = F p X/F p+1 X hence exactness
in the middle. The two short exact sequence of (2) are special cases of the short
exact sequence of (1).

Lemma 16.13. Let A be an abelian category. Let f : A B be a morphism of
finite filtered objects of A. The following are equivalent
(1)
(2)
(3)
(4)
(5)
(6)
f is strict,
the morphism Coim(f ) Im(f ) is an isomorphism,
gr(Coim(f )) gr(Im(f )) is an isomorphism,
the sequence gr(Ker(f )) gr(A) gr(B) is exact,
the sequence gr(A) gr(B) gr(Coker(f )) is exact, and
the sequence
0 gr(Ker(f )) gr(A) gr(B) gr(Coker(f )) 0
is exact.
Proof. The equivalence of (1) and (2) is Lemma 16.4. By Lemma 16.12 we see that
(4), (5), (6) imply (3) and that (3) implies (4), (5), (6). Hence it suffices to show that
(3) implies (2). Thus we have to show that if f : A B is an injective and surjective
map of finite filtered objects which induces and isomorphism gr(A) gr(B), then
f induces an isomorphism of filtered objects. In other words, we have to show
that f (F p A) = F p B for all p. As the filtrations are finite we may prove this by
HOMOLOGICAL ALGEBRA
39
descending induction on p. Suppose that f (F p+1 A) = F p+1 B. Then commutative

diagram
/ F p+1 A
/ F p+1 B
/ F pA
f
/ F pB
/ grp (A)
/0
grp (f )

/ grp (B)
/0
and the five lemma imply that f (F p A) = F p B.
Lemma 16.14. Let A be an abelian category. Let A B C be a complex of

filtered objects of A. Assume : A B and : B C are strict morphisms of
filtered objects. Then gr(Ker()/Im()) = Ker(gr())/Im(gr())).
Proof. This follows formally from Lemma 16.12 and the fact that Coim()
=
Im() and Coim()

= Im() by Lemma 16.4.
Lemma 16.15. Let A be an abelian category. Let A B C be a complex
of filtered objects of A. Assume A, B, C have finite filtrations and that gr(A)
gr(B) gr(C) is exact. Then
(1)
(2)
(3)
(4)
(5)
for each p Z the sequence grp (A) grp (B) grp (C) is exact,
for each p Z the sequence F p (A) F p (B) F p (C) is exact,
for each p Z the sequence A/F p (A) B/F p (B) C/F p (C) is exact,
the maps A B and B C are strict, and
A B C is exact (as a sequence in A).
Proof. Part (1) is immediate from the definitions. We will prove (3) by induction
on the length of the filtrations. If each of A, B, C has only one nonzero graded part,
then (3) holds as gr(A) = A, etc. Let n be the largest integer such that at least
one of F n A, F n B, F n C is nonzero. Set A0 = A/F n A, B 0 = B/F n B, C 0 = C/F n C
with induced filtrations. Note that gr(A) = F n A gr(A0 ) and similarly for B
and C. The induction hypothesis applies to A0 B 0 C 0 , which implies that
A/F p (A) B/F p (B) C/F p (C) is exact for p n. To conclude the same for
p = n + 1, i.e., to prove that A B C is exact we use the commutative diagram
0
/ F nA
/A
/ A0
/0

/ F nB

/B

/ B0
/0

/ F nC

/C

/ C0
/0
whose rows are short exact sequences of objects of A. The proof of (2) is dual. Of
course (5) follows from (2).
To prove (4) denote f : A B and g : B C the given morphisms. We know
that f (F p (A)) = Ker(F p (B) F p (C)) by (2) and f (A) = Ker(g) by (5). Hence
f (F p (A)) = Ker(F p (B) F p (C)) = Ker(g) F p (B) = f (A) F p (B) which proves
that f is strict. The proof that g is strict is dual to this.
40
HOMOLOGICAL ALGEBRA
17. Spectral sequences

A nice discussion of spectral sequences may be found in [Eis95]. See also [McC01],
[Lan02], etc.
(1) A spectral sequence in A is given by a system (Er , dr )r1 where each Er is
an object of A, each dr : Er Er is a morphism such that dr dr = 0 and
Er+1 = Ker(dr )/Im(dr ) for r 1.
(2) A morphism of spectral sequences f : (Er , dr )r1 (Er0 , d0r )r1 is given by
a family of morphisms fr : Er Er0 such that fr dr = d0r fr and such
that fr+1 is the morphism induced by fr via the identifications Er+1 =
0
Ker(dr )/Im(dr ) and Er+1
= Ker(d0r )/Im(d0r ).
We will sometimes loosen this definition somewhat and allow Er+1 to be an object
with a given isomorphism Er+1 Ker(dr )/Im(dr ). In addition we sometimes
have a system (Er , dr )rr0 for some r0 satisfying the properties of the definition
above for indices r. We will also call this a spectral sequence since by a simple
renumbering it falls under the definition anyway. In fact, sometimes it makes sense
to allow r0 = 0 or even r0 = 1 due to conventions in the literature.
Given a spectral sequence (Er , dr )r1 we define
0 = B1 B2 . . . Br . . . Zr . . . Z2 Z1 = E1
by the following simple procedure. Set B2 = Im(d1 ) and Z2 = Ker(d1 ). Then it is
clear that d2 : Z2 /B2 Z2 /B2 . Hence we can define B3 as the unique subobject
of E1 containing B2 such that B3 /B2 is the image of d2 . Similarly we can define
Z3 as the unique subobject of E1 containing B2 such that Z3 /B2 is the kernel of
d2 . And so on and so forth. In particular we have
Er = Zr /Br
for all r 1. I case the spectral sequence starts at r = r0 then we can similarly
construct Bi , Zi as subobjects in Er0 .
Definition 17.2. Let A be an abelian category. Let (Er , dr )r1 be a spectral
sequence.
T
S
(1) If the subobjects Z = Zr and B = Br of E1 exist then we define
the limit of the spectral sequence to be the object
E = Z /B .
(2) We say that the spectral sequence collapses at Er , or degenerates at Er if
the differentials dr , dr+1 , . . . are all zero.
Note that if the spectral sequence collapses at Er , then we have Er = Er+1 =
. . . = E (and the limit exists of course). Also, almost any abelian category we
will encounter has countable sums and intersections.
Remark 17.3 (Variant). It is often the case that the terms of a spectral sequence
have additional structure, for example a grading or a bigrading. To accomodate
this (and to get around certain technical issues) we introduce the following notion.
Let A be an abelian category. Let (Tr )r1 be a sequence of translation or shift
functors, i.e., Tr : A A is an isomorphism of categories. In this setting a spectral
HOMOLOGICAL ALGEBRA
41
sequence is given by a system (Er , dr )r1 where each Er is an object of A, each

dr : Er Tr Er is a morphism such that Tr dr dr = 0 so that
...
/ Tr1 Er
Tr1 dr
/ Er
/ Tr Er
dr
Tr dr
/ Tr2 Er
/ ...
is a complex and Er+1 = Ker(dr )/Im(Tr1 dr ) for r 1. It is clear what a morphism

of spectral sequences means in this setting. In this setting we can still define
0 = B1 B2 . . . Br . . . Zr . . . Z2 Z1 = E1
and Z and B (if they exist) as above.
18. Spectral sequences: exact couples
(1) An exact couple is a datum (A, E, , f, g) where A, E are objects of A and
, f , g are morphisms as in the following diagram
A_
/A
with the property that the kernel of each arrow is the image of its predecessor. So Ker() = Im(f ), Ker(f ) = Im(g), and Ker(g) = Im().
(2) A morphism of exact couples t : (A, E, , f, g) (A0 , E 0 , 0 , f 0 , g 0 ) is given
by morphisms tA : A A0 and tE : E E 0 such that 0 tA = tA ,
f 0 tE = tA f , and g 0 tA = tE g.
Lemma 18.2. Let (A, E, , f, g) be an exact couple in an abelian category A. Set
(1)
(2)
(3)
(4)
(5)
(6)
d = g f : E E so that d d = 0,
E 0 = Ker(d)/Im(d),
A0 = Im(),
0 : A0 A0 induced by ,
f 0 : E 0 A0 induced by f ,
g 0 : A0 E 0 induced by g 1 .
Then we have
(1) Ker(d) = f 1 (Ker(g)) = f 1 (Im()),
(2) Im(d) = g(Im(f )) = g(Ker()),
(3) (A0 , E 0 , 0 , f 0 , g 0 ) is an exact couple.
Proof. Omitted.
Hence it is clear that given an exact couple (A, E, , f, g) we get a spectral sequence
by setting E1 = E, d1 = d, E2 = E 0 , d2 = d0 = g 0 f 0 , E3 = E 00 , d3 = d00 = g 00 f 00 ,
and so on.
Definition 18.3. Let A be an abelian category. Let (A, E, , f, g) be an exact
couple. The spectral sequence associated to the exact couple is the spectral sequence
(Er , dr )r1 with E1 = E, d1 = d, E2 = E 0 , d2 = d0 = g 0 f 0 , E3 = E 00 , d3 = d00 =
g 00 f 00 , and so on.
42
HOMOLOGICAL ALGEBRA
Lemma 18.4. Let A be an abelian category. Let (A, E, , f, g) be an exact couple.

Let (Er , dr )r1 be the spectral sequence associated to the exact couple. In this case
we have
0 = B1 . . . Br+1 = g(Ker(r )) . . . Zr+1 = f 1 (Im(r )) . . . Z1 = E
and the map dr+1 : Er+1 Er+1 is described by the following rule: For any (test)
object T of A and any elements x : T Zr+1 and y : T A such that f x = r y
we have
dr+1 x = g y
where x : T Er+1 is the induced morphism.
Proof. Omitted.
Note that in the situation of the lemma we obviously have

[
\
Ker(r ) Z = f 1
Im(r )
B = g
r
provided these exist and in this case E = Z /B .

Remark 18.5 (Variant). Let A be an abelian category. Let S, T : A A be shift
functors, i.e., isomorphisms of categories. We will indicate the n-fold compositions
by S n A and T n A for A Ob(A) and n Z. In this situation an exact couple is a
datum (A, E, , f, g) where A, E are objects of A and : A T 1 A, f : E A,
g : A SE are morphisms such that
TE
Tf
/ TA
/A
/ SE
Sf
/ SA
is an exact complex. Lets visualize this as follows

T Aa
/A
_
T
Tf
TE
SE
y
T 1 SE
/ T 1 A
T 1 g
We set d = g f : E SE. Then d S 1 d = g f S 1 g S 1 f = 0 because

f S 1 g = 0. Set E 0 = Ker(d)/Im(S 1 d). Set A0 = Im(T ). Let 0 : A0 T 1 A0
induced by . Let f 0 : E 0 A0 be induced by f which works because f (Ker(d))
Ker(g) = Im(T ). Finally, let g 0 : A0 T SE 0 induced by T g (T )1 3.
In exactly the same way as above we find
(1) Ker(d) = f 1 (Ker(g)) = f 1 (Im(T )),
(2) Im(d) = g(Im(f )) = g(Ker()),
(3) (A0 , E 0 , 0 , f 0 , g 0 ) is an exact couple for the shift functors T S and T .
We obtain a spectral sequence (as in Remark 17.3) with E1 = E, E2 = E 0 , etc,
with dr : Er T r1 SEr for all r 1. Lemma 18.4 tells us that
SBr+1 = g(Ker(T r+1 . . . T 1 ))
and
Zr+1 = f 1 (Im(T T 2 . . . T r ))
3This works because T SE 0 = Ker(T Sd)/Im(T d) and T g(Ker(T )) = T g(Im(T f )) = Im(T (d))
and T S(d)(Im(T g)) = Im(T Sg T Sf T g) = 0.
HOMOLOGICAL ALGEBRA
43
in this situation. The description of the map dr+1 is similar to that given in the
lemma. (It may be easier to use these explicit descriptions to prove one gets a
spectral sequence from such an exact couple.)
19. Spectral sequences: differential objects
Definition 19.1. Let A be an abelian category. A differential object of A is a pair
(A, d) consisting of an object A of A endowed with a selfmap d such that dd = 0. A
morphism of differential objects (A, d) (B, d) is given by a morphism : A B
such that d = d.
Lemma 19.2. Let A be an abelian category. The category of differential objects of
A is abelian.
Proof. Omitted.
Definition 19.3. For a differential object (A, d) we denote

H(A, d) = Ker(d)/Im(d)
its homology.
Lemma 19.4. Let A be an abelian category. Let 0 (A, d) (B, d) (C, d) 0
be a short exact sequence of differential objects. Then we get an exact homology
sequence
. . . H(C, d) H(A, d) H(B, d) H(C, d) . . .
Proof. Apply Lemma 12.12 to the short exact sequence of complexes
0
0
0
where the vertical arrows are d.
We come to an important example of a spectral sequence. Let A be an abelian

category. Let (A, d) be a differential object of A. Let : (A, d) (A, d) be an
endomorphism of this differential object. If we assume injective, then we get a
short exact sequence
0 (A, d) (A, d) (A/A, d) 0
of differential objects. By the Lemma 19.4 we get an exact couple
H(A, d)
f
/ H(A, d)
x
H(A/A, d)
where g is the canonical map and f is the map defined in the snake lemma. Thus we
get an associated spectral sequence! Since in this case we have E1 = H(A/A, d)
we see that it makes sense to define E0 = A/A and d0 = d. In other words, we
start the spectral sequence with r = 0. According to our conventions in Section 17
we define a sequence of subobjects
0 = B0 . . . Br . . . Zr . . . Z0 = E0
44
HOMOLOGICAL ALGEBRA
with the property that Er = Zr /Br . Namely we have for r 1 that

(1) Br is the image of (r1 )1 (dA) under the natural map A A/A,
(2) Zr is the image of d1 (r A) under the natural map A A/A, and
(3) dr : Er Er is given as follows: given an element z Zr choose an element
y A such that d(z) = r (y). Then dr (z + Br + A) = y + Br + A.
Warning: It is not necessarily the case that A (r1 )1 (dA), nor A
d1 (r A). It is true that (r1 )1 (dA) d1 (r A). We have
Er =
d1 (r A) + A
.
(r1 )1 (dA) + A
It is not hard to verify directly that (1) (3) give a spectral sequence.
Definition 19.5. Let A be an abelian category. Let (A, d) be a differential object
of A. Let : A A be an injective selfmap of A which commutes with d.
The spectral sequence associated to (A, d, ) is the spectral sequence (Er , dr )r0
described above.
Remark 19.6 (Variant). Let A be an abelian category and let S, T : A A be
shift functors, i.e., isomorphisms of categories. Assume that T S = ST as functors.
Consider pairs (A, d) consisting of an object A of A and a morphism d : A SA
such that d S 1 d = 0. The category of these objects is abelian. We define
H(A, d) = Ker(d)/Im(S 1 d) and we observe that H(SA, Sd) = SH(A, d) (canonical isomorphism). Given a short exact sequence
0 (A, d) (B, d) (C, d) 0
we obtain a long exact homology sequence
. . . S 1 H(C, d) H(A, d) H(B, d) H(C, d) SH(A, d) . . .
(note the shifts in the boundary maps). Since ST = T S the functor T defines a shift
functor on pairs by setting T (A, d) = (T A, T d). Next, let : (A, d) T 1 (A, d)
be injective with cokernel (Q, d). Then we get an exact couple as in Remark 18.5
with shift functors T S and T given by
(H(A, d), S 1 H(Q, d), , f, g)
where : H(A, d) T 1 H(A, d) is induced by , the map f : S 1 H(Q, d)
H(A, d) is the boundary map and g : H(A, d) T H(Q, d) = T S(S 1 H(Q, d))
is induced by the quotient map A T Q. Thus we get a spectral sequence as
above with E1 = S 1 H(Q, d) and differentials dr : Er T r SEr . As above we set
E0 = S 1 Q and d0 : E0 SE0 given by S 1 d : S 1 Q Q. If according to our
conventions we define Br Zr E0 , then we have for r 1 that
(1) SBr is the image of
(T r+1 . . . T 1 )1 Im(T r S 1 d)
under the natural map T 1 A Q,
(2) Zr is the image of
(S 1 T 1 d)1 Im( . . . T r1 )
under the natural map S 1 T 1 A S 1 Q.
HOMOLOGICAL ALGEBRA
45
The differentials can be described as follows: if x Zr , then pick x0 S 1 T 1 A

mapping to x. Then S 1 T 1 d(x0 ) is ( . . . T r1 )(y) for some y T r1 A.
Then dr (x) T r SEr is represented by the class of the image of y in T r SE0 = T r Q
modulo T r SBr .
20. Spectral sequences: filtered differential objects
We can build a spectral sequence starting with a filtered differential object.
Definition 20.1. Let A be an abelian category. A filtered differential object
(K, F, d) is a filtered object (K, F ) of A endowed with an endomorphism d :
(K, F ) (K, F ) whose square is zero: d d = 0.
To describe the spectral sequence associated to such an object we assume, for
the moment, that A is an abelian category which has countable direct sums and
countable direct sums are exact (this is not automatic, see Remark 15.3). Let
(K, F, d) be a filtered differential object ofL
A. Note that each F n K is a differential
object by itself. Consider the object A =
F n K and endow it with a differential
d by using d on each summand. Then (A, d) is a differential object of A which
comes equipped with a grading. Consider the map
:AA
n
which is given by the inclusions F A F n1 A. This is clearly an injective morphism of differential objects : (A, d) (A, d). Hence, by Definition 19.5 we get
a spectral sequence. We will call this the spectral sequence associated to the filtered
differential object (K, F, d).
Let us figure out the terms of this spectral sequence. First, note that A/A = gr(K)
endowed with its differential d = gr(d). Hence we see that
E0 = gr(K),
d0 = gr(d).
Hence the homology of the graded differential object gr(K) is the next term:
E1 = H(gr(K), gr(d)).
In addition we see that E0 is a graded object of A and that d0 is compatible with
the grading. Hence clearly E1 is a graded object as well. But it turns out that the
differential d1 does not preserve this grading; instead it shifts the degree by 1.
To work this out precisely, we define
Zrp =
F p K d1 (F p+r K) + F p+1 K
F p+1 K
and
F p K d(F pr+1 K) + F p+1 K
.
F p+1 K
This notation, although quite natural, seems to be different from the notation in
most places in the literature. Perhaps it does not matter, since the literature does
not seem to have a consistent choice of notation either. With these choices
we see
L
that Br E0 , resp. Zr E0 (as defined in Section 19) is equal to p Brp , resp.
L p
p Zr . Hence if we define
Erp = Zrp /Brp
Brp =
46
HOMOLOGICAL ALGEBRA
for r 0 and p Z, then we have Er =

dpr : Erp Erp+r by the rule
Erp . We can define a differential
z + F p+1 K 7 dz + F p+r+1 K
where z F p K d1 (F p+r K).
Lemma 20.2. Let A be an abelian category. Let (K, F, d) be a filtered differential
object of A. There is a spectral sequence (Er , dr )r0 in Gr(A) associated to (K, F, d)
such that dr : Er Er [1] for all r and such that the graded pieces Erp and maps
dpr : Erp Erp+r are as given above. Furthermore, E0p = grp K, dp0 = grp (d), and
E1p = H(grp K, d).
Proof. If A has countable direct sums and if countable direct sums are exact, then
this follows from the discussion above. In general, we proceed as follows; we strongly
suggest the reader skip this proof. Consider the object A = (F p+1 K) of Gr(A), i.e.,
we put F p+1 K in degree p (the funny shift in numbering to get numbering correct
later on). We endow it with a differential d by using d on each component. Then
(A, d) is a differential object of Gr(A). Consider the map
: A A[1]
which is given in degree p by the inclusions F p+1 A F p A. This is clearly an
injective morphism of differential objects : (A, d) (A, d)[1]. Hence, we can
apply Remark 19.6 with S = id and T = [1]. The corresponding spectral sequence
(Er , dr )r0 in Gr(A) is the spectral sequence we are looking for. Let us unwind the
definitions a bit. First of all we have Er = (Erp ) is an object of Gr(A). Then, since
T r S = [r] we have dr : Er Er [r] which means that dpr : Erp Erp+r .
To see that the description of the graded pieces hold, we argue as above. Namely,
first we have E0 = Coker( : A A[1]) and by our choice of numbering above
this gives E0p = grp K. The first differential is given by dp0 = grp d : E0p E0p . Next,
the description of the boundaries Br and the cocycles Zr in Remark 19.6 translates

into a straightforward manner into the formulae for Zrp and Brp given above.
object of A. The spectral sequence (Er , dr )r0 associated to (K, F, d) has
dp1 : E1p = H(grp K) H(grp+1 K) = E1p+1
equal to the boundary map in homology associated to the short exact sequence of
differential objects
0 grp+1 K F p K/F p+2 K grp+1 K 0.
Proof. Omitted.
Definition 20.4. Let A be an abelian category. Let (K, F, d) be a filtered differential object of A. The induced filtration on H(K, d) is the filtration defined by
F p H(K, d) = Im(H(F p K, d) H(K, d)).
p
p
object of A. If Z
and B
exist (see proof ), then associated graded gr(H(K))
of the cohomology of K is a graded subquotient of the graded object E having
p
p
p
E
= Z
/B
in degree p.
HOMOLOGICAL ALGEBRA
47
Proof. Here we have

p
Z
=
p
B
Zrp =
and
=
Brp
S
=
r (F
K d1 (F p+r K) + F p+1 K)
F p+1 K
r (F
K d(F pr+1 K) + F p+1 K)

.
F p+1 K
Thus
T
(F p K d1 (F p+r K) + F p+1 K)
p
E
= Sr p
.
pr+1 K) + F p+1 K)
r (F K d(F
and the top and bottom exist. On the other hand, we have
grp H(K) =
Ker(d) F p K + F p+1 K
Im(d) F p K + F p+1 K
The result follows since

(20.5.1)
Ker(d) F p K + F p+1 K
[
r
F p K d1 (F p+r K) + F p+1 K
and
(20.5.2)
\
r

F p K d(F pr+1 K) + F p+1 K Im(d) F p K + F p+1 K.
Definition 20.6. Let A be an abelian category. Let (K, F, d) be a filtered differential object of A. We say the spectral sequence associated to (K, F, d) converges
if gr(H(K)) = E via Lemma 20.5. In this case we also say that (Er , dr )r0 abuts
to or converges to H(K).
In the literature one finds more refined notions distinguishing between weakly
converging, abutting and converging. Namely, one can require the filtration
on H(K) to be either arbitrary, or exhaustive and separated, or exhaustive
and complete in addition to the condition that gr(H(K)) = E . We try to
avoid introducing this notation by simply adding the relevant information in the
statements of the results.
object of A. The associated spectral sequence converges if and only if for every
p Z we have equality in equations (20.5.2) and (20.5.1).
Proof. Immediate from the discussions above.
21. Spectral sequences: filtered complexes

Definition 21.1. Let A be an abelian category. A filtered complex K of A is a
complex of Fil(A) (see Definition 16.1).
We will denote the filtration on the objects by F . Thus F p K n denotes the pth step
in the filtration of the nth term of the complex. Note that each F p K is a complex
of A. Hence we could also have defined a filtered complex as a filtered object in
the (abelian) category of complexes of A. In particular grK is a graded object of
the category of complexes of A.
To describe the spectral sequence associated to such an object we assume, for
the moment, that A is an abelian category which has countable direct sums and
countable direct sums are exact (this is not automatic, see Remark 15.3). Let
48
HOMOLOGICAL ALGEBRA
L n
us denote d the differential of K. Forgetting the grading we can think of
K
as a filtered differential object of A. Hence according to Section 20 we obtain
a spectral sequence (Er , dr )r0 . In this section we work out the terms of this
spectral sequence, and we endow the terms of this spectral sequence with additional
structure coming from the grading of K.
First we point out that E0p = grp K is a complex and hence is graded. Thus E0 is
bigraded in a natural way. It is customary to use the bigrading
M
E0 =
E0p,q , E0p,q = grp K p+q
p,q
The idea is that p + q should be thought of as the total degree of the (co)homology
classes. Also, p is called the filtration degree, and q is called the complementary
degree. The differential d0 is compatible with this bigrading in the following way
M p,q
p,q
p,q+1
d0 , dp,q
.
d0 =
0 : E0 E0
Namely, dp0 is just the differential on the complex grp K (which occurs as grp E0
just shifted a bit).
To go further we identify the objects Brp and Zrp introduced in Section 20 as graded
objects and we work out the corresponding decompositions of the differentials. We
do this in a completely straightforward manner, but again we warn the reader that
our notation is not the same as notation found elsewhere. We define
Zrp,q =
F p K p+q d1 (F p+r K p+q+1 ) + F p+1 K p+q

F p+1 K p+q
and
F p K p+q d(F pr+1 K p+q1 ) + F p+1 K p+q
.
F p+1 K p+q
and ofL
course Erp,q = L
Zrp,q /Brp,q . With these
L definitions it is completely clear that
p
p
p,q
Zr = q Zr , Br = q Brp,q , and Erp = q Erp,q . Moreover,
Brp,q =
0 . . . Brp,q . . . Zrp,q . . . E0p,q

S
T
p,q
p,q
p,q
=
= r Brp,q and E
= r Zrp,q and B
and hence it makes sense to define Z
p,q
p
p,q
Z /B provided these exist. Also, the map dr decomposes as the direct sum of
the maps
p,q
dp,q
Erp+r,qr+1 ,
r : Er
z + F p+1 K p+q 7 dz + F p+r+1 K p+q+1
where z F p K p+q d1 (F p+r K p+q+1 ).

Lemma 21.2. Let A be an abelian category. Let (K , F ) be a filtered complex of
A. There is a spectral sequence (Er , dr )r0 in the category of bigraded objects of A
associated to (K , F ) such that dr has bidegree (r, r + 1) and such that Er has bip,q
graded pieces Erp,q and maps dp,q
Erp+r,qr+1 as given above. Furthermore,
r : Er
p,q
p,q
p
p+q
p p+q
), d0 = gr (d ), and E1p,q = H p+q (grp (K )).
we have E0 = gr (K
Proof. If A has countable direct sums and if countable direct sums are exact, then
this follows from the discussion above. In general, we proceed as follows; we strongly
suggest the reader skip this proof. Consider the bigraded object A = (F p+1 K p+1+q )
of A, i.e., we put F p+1 K p+1+q in degree (p, q) (the funny shift in numbering to get
numbering correct later on). We endow it with a differential d : A A[0, 1] by
HOMOLOGICAL ALGEBRA
49
using d on each component. Then (A, d) is a differential bigraded object. Consider

the map
: A A[1, 1]
which is given in degree (p, q) by the inclusion F p+1 K p+q F p K p+q . This is
an injective morphism of differential objects : (A, d) (A, d)[1, 1]. Hence,
we can apply Remark 19.6 with S = [0, 1] and T = [1, 1]. The corresponding
spectral sequence (Er , dr )r0 of bigraded objects is the spectral sequence we are
looking for. Let us unwind the definitions a bit. First of all we have Er = (Erp,q ).
Then, since T r S = [r, r + 1 we have dr : Er Er [r, r + 1] which means that
dpr : Erp,q Erp+r,qr+1 .
To see that the description of the graded pieces hold, we argue as above. Namely,
first we have
E0 = Coker( : A A[1, 1])[0, 1] = Coker([0, 1] : A[0, 1] A[1, 0])
and by our choice of numbering above this gives
E0p,q = Coker(F p+1 K p+q F p K p+q ) = grp K p+q
The first differential is given by dp,q
= grp dp+q : E0p,q E0p,q+1 . Next, the de0
scription of the boundaries Br and the cocycles Zr in Remark 19.6 translates into
a straightforward manner into the formulae for Zrp,q and Brp,q given above.

A. Assume A has countable direct sums. Let (Er , dr )r0 be the spectral sequence
associated to (K , F ).
(1) The map
p,q
p+q
dp,q
(grp (K )) E1p+1,q = H p+q+1 (grp+1 (K ))
1 : E1 = H
is equal to the boundary map in cohomology associated to the short exact

sequence of complexes
0 grp+1 (K ) F p K /F p+2 K grp+1 (K ) 0.
(2) Assume that d(F p K) F p+1 K for all p Z. Then d induces the zero
differential on grp (K ) and hence E1p,q = grp (K )p+q . Furthermore, in this
case
p,q
p
p+q
dp,q
E1p,q = grp+1 (K )p+q+1
1 : E1 = gr (K )
is the morphism induced by d.

Proof. Omitted. But compare Lemma 20.3.
Lemma 21.4. Let A be an abelian category. Let : (K , F ) (L , F ) be a

morphism of filtered complexes of A. Let (Er (K), dr )r0 , resp. (Er (L), dr )r0 be
the spectral sequence associated to (K , F ), resp. (L , F ). The morphism induces
a canonical morphism of spectral sequences {r : Er (K) Er (L)}r0 compatible
with the bigradings.
Proof. Obvious from the explicit representation of the terms of the spectral sequences.

Definition 21.5. Let A be an abelian category. Let (K , F ) be a filtered complex
of A. The induced filtration on H n (K ) is the filtration defined by F p H n (K ) =
Im(H n (F p K ) H n (K )).
50
HOMOLOGICAL ALGEBRA

p,q
p,q
A. If Z
and B
exist (see above), then the associated gradedL
gr(H n (K )) of
p,q
the cohomology of K is a graded subquotient of the graded object p+q=n E
.
Proof. Let q = n p. As in the proof of Lemma 20.5 we see that
T
(F p K n d1 (F p+r K n+1 ) + F p+1 K n )
p,q
E
= Sr p n
.
pr+1 K n1 ) + F p+1 K n )
r (F K d(F
On the other hand, we have
grp H n (K) =
(21.6.1)
Ker(d) F p K n + F p+1 K n
Im(d) F p K n + F p+1 K n
The result follows since

(21.6.2) Ker(d) F p K n + F p+1 K n
[
r
F p K n d1 (F p+r K n+1 ) + F p+1 K n
and
(21.6.3)
\
r

F p K n d(F pr+1 K n1 ) + F p+1 K n Im(d) F p K n + F p+1 K n .
Definition 21.7. Let A be an abelian category. Let (K , F ) be a filtered complex

of
the spectral sequence associated to (K , F ) converges if grH n (K ) =
LA. We say
p,q
p+q=n E for every n Z.
This is often symbolized by the notation Erp,q H p+q (K ). Please read the
remarks following Definition 20.6.
A. The associated spectral sequence converges if and only if for every p, q Z we
have equality in equations (21.6.3) and (21.6.2).
Proof. Immediate from the discussions above.

A. Assume that the filtration on each K n is finite (see Definition 16.1). Then
(1) the filtration on each H n (K ) is finite, and
(2) the spectral sequence associated to (K , F ) converges.
Proof. Part (1) is clear from Equation (21.6.1). We will use Lemma 21.8 to prove
part (2). Fix p, n Z. Look at the left hand side of Equation (21.6.3). The
expression is equal to the right hand side since F m K n1 = 0 for m 0. Similarly,
use F m K n+1 = K n+1 for m 0 to prove equality in Equation (21.6.2).

22. Spectral sequences: double complexes
Definition 22.1. Let A be an additive category. A double complex in A is given
p,q
p,q
by a system ({Ap,q , dp,q
is an object of A and dp,q
:
1
1 , d2 }p,qZ ), where each A
p,q
Ap,q Ap+1,q and d2 : Ap,q Ap,q+1 are morphisms of A such that the following
rules hold:
(1) dp+1,q
dp,q
1 =0
1
p,q+1
(2) d2
dp,q
2 =0
p,q
p+1,q
(3) dp,q+1
d
dp,q
1
2 = d2
1
for all p, q Z.
HOMOLOGICAL ALGEBRA
51
This is just the cochain version of the definition. It says that each Ap, is a cochain
complex and that each dp,
is a morphism of complexes Ap, Ap+1, such that
1
p,
p+1,
d1 = 0 as morphisms of complexes. In other words a double complex can
d1
be seen as a complex of complexes. So in the diagram
...
. .O .
...
/ Ap,q+1
O
. .O .
...
/ Ap+1,q+1
O
/ ...
dp,q+1
1
dp,q
2
dp+1,q
2
/ Ap,q
O
...
dp,q
1
/ Ap+1,q
O
/ ...
...
...
...
...
any square commutes. Warning: In the literature one encounters a different definition where a bicomplex or a double complex has the property that the squares
in the diagram anti-commute.
Example 22.2. Let A, B, C be abelian categories. Suppose that
: A B C,
(X, Y ) 7 X Y
is a functor which is bilinear on morphisms, see Categories, Definition 2.20 for the
definition of A B. Given a complexes X of A and Y of B we obtain a double
complex
K , = X Y
in C. Here the first differential K p,q K p+1,q is the morphism X p Y q
X p+1 Y q induced by the morphism X p X p+1 and the identity on Y q . Similarly
for the second differential.
Let A, be a double complex. It is customary to denote HIp (A, ) the complex
p1,q
with terms Ker(dp,q
) (varying q) and differential induced by d2 . Then
1 )/Im(d1
p
q
,
HII (HI (A )) denotes its cohomology in degree q. It is also customary to denote
p,q1
q
(A, ) the complex with terms Ker(dp,q
) (varying p) and differential
HII
2 )/Im(d2
p
q
,
induced by d1 . Then HI (HII (A )) denotes its cohomology in degree q. It will turn
out that these cohomology groups show up as the terms in the spectral sequence
for a filtration on the associated to total complex.
Definition 22.3. Let A be an additive category. Let A, be a double complex.
The associated simple complex sA , also sometimes called the associated total complex is given by
M
sAn =
Ap,q
n=p+q
(if it exists) with differential

dnsA =
X
n=p+q
p p,q
(dp,q
1 + (1) d2 )
Alternatively, we sometimes write Tot(A, ) to denote this complex.

If countable direct sums exist in A or if for each n at most finitely many Ap,np are
nonzero, then sA exists. Note that the definition is not symmetric in the indices
(p, q).
52
HOMOLOGICAL ALGEBRA
There are two natural filtrations on the simple complex sA associated to the double
complex A, . Namely, we define
M
M
p
FIp (sAn ) =
Ai,j and FII
(sAn ) =
Ai,j .
i+j=n, ip
i+j=n, jp
It is immediately verified that (sA , FI ) and (sA , FII ) are filtered complexes. By
Section 21 we obtain two spectral sequences. It is customary to denote (0 Er , 0 dr )r0
the spectral sequence associated to the filtration FI and to denote (00 Er , 00 dr )r0
the spectral sequence associated to the filtration FII . Here is a description of these
spectral sequences.
Lemma 22.4. Let A be an abelian category. Let K , be a double complex. The
spectral sequences associated to K , have the following terms:
p,q
p p,q
K p,q+1 ,
(1) 0 E0p,q = K p,q with 0 dp,q
0 = (1) d2 : K
p,q
p,q
q,p
(2) 00 E0 = K q,p with 00 d0 = d1 : K q,p K q+1,p ,
q p,
(3) 0 E1p,q = H q (K p, ) with 0 dp,q
1 = H (d1 ),
p,q
p,q
(4) 00 E1 = H q (K ,p ) with 00 d1 = (1)q H q (d,p
2 ),
q
(5) 0 E2p,q = HIp (HII
(K , )),
p
(6) 00 E2p,q = HII
(HIq (K , )).
Proof. Omitted.
These spectral sequences define two filtrations on H n (sK ). We will denote these
FI and FII .
Definition 22.5. Let A be an abelian category. Let K , be a double complex.
We say the spectral sequence (0 Er , 0 dr )r0 converges if Definition 21.7 applies. In
other words, for all n
p,q
grFI (H n (sK )) = p+q=n 0 E
via the canonical comparison of Lemma 21.6. Similarly we say the spectral sequence
(00 Er , 00 dr )r0 converges if Definition 21.7 applies. In other words for all n
p,q
grFII (H n (sK )) = p+q=n 00 E
via the canonical comparison of Lemma 21.6.

Same caveats as those following Definition 20.6.
Lemma 22.6 (First quadrant spectral sequence). Let A be an abelian category.
Let K , be a double complex. Assume that for every n Z there are only finitely
many nonzero K p,q with p + q = n. Then
(1) the filtrations FI , FII on each H n (K ) are finite,
(2) the spectral sequence (0 Er , 0 dr )r0 converges, and
(3) the spectral sequence (00 Er , 00 dr )r0 converges.
Proof. Follows immediately from Lemma 21.9.
Here is our first application of spectral sequences.

Lemma 22.7. Let A be an abelian category. Let K be a complex. Let A, be a
double complex. Let p : K p Ap,0 be morphisms. Assume that
(1) For every n Z there are only finitely many nonzero Ap,q with p + q = n.
(2) We have Ap,q = 0 if q < 0.
(3) The morphisms p give rise to a morphism of complexes : K A,0 .
HOMOLOGICAL ALGEBRA
53
(4) The complex Ap, is exact in all degrees q 6= 0 and the morphism K p Ap,0
induces an isomorphism K p Ker(dp,0
2 ).
Then induces a quasi-isomorphism
K sA
of complexes. Moreover, there is a variant of this lemma involving the second
variable q instead of p.
Proof. The map is simply the map given by the morphisms K n An,0 sAn ,
which are easily seen to define a morphism of complexes. Consider the spectral
sequence (0 Er , 0 dr )r0 associated to the double complex A, . By Lemma 22.6 this
spectral sequence converges and the induced filtration on H n (sA ) is finite for each
n. By Lemma 22.4 and assumption (4) we have 0 E1p,q = 0 unless q = 0 and 0 E1p,0 =
K p with differential 0 dp,0
identified with dpK . Hence 0 E2p,0 = H p (K ) and zero
1
p,q
otherwise. This clearly implies dp,q
2 = d3 = . . . = 0 for degree reasons. Hence we
n
conclude that H (sA ) = H (K ). We omit the verification that this identification

is given by the morphism of complexes K sA introduced above.

Remark 22.8. Let A be an abelian category. Let C A be a weak Serre subcategory (see Definition 9.1). Suppose that K , is a double complex to which Lemma
22.6 applies such that for some r 0 all the objects 0 Erp,q belong to C. We claim
all the cohomology groups H n (sK ) belong to C. Namely, the assumptions imply
0 p,q
that the kernels and images of 0 dp,q
r are in C. Whereupon we see that each Er+1 is
n
0 p,q
in C. By induction we see that each E is in C. Hence each H (sK ) has a finite
filtration whose subquotients are in C. Using that C is closed under extensions we
conclude that H n (sK ) is in C as claimed.
The same result holds for the second spectral sequence associated to K , . Similarly,
if (K , F ) is a filtered complex to which Lemma 21.9 applies and for some r 0
all the objects Erp,q belong to C, then each H n (K ) is an object of C.
Remark 22.9. Let A be an additive category. Let A,, be a triple complex. The
associated total complex is the complex with terms
M
Totn (A,, ) =
Ap,q,r
p+q+r=n
and differential
dnTot(A,, ) =
X
p+q+r=n
+ (1)p+q dp,q,r
dp,q,r
+ (1)p dp,q,r
1
2
3
With this definition a simple calculation shows that the associated total complex
is equal to
Tot(A,, ) = Tot(Tot12 (A,, )) = Tot(Tot23 (A,, ))
In other words, we can either first combine the first two of the variables and then
combine sum of those with the last, or we can first combine the last two variables
and then combine the first with the sum of the last two.
Lemma 22.10. Let M be a complex of abelian groups. Let
. . . A2 A1 A0 M 0
54
HOMOLOGICAL ALGEBRA
be an exact complex of complexes of abelian groups such that for all p Z the
complexes
. . . Ker(dpA ) Ker(dpA ) Ker(dpA ) Ker(dpM ) 0
2
Aqp
are exact as well. Set A =

to obtain a double complex. Then Tot(A, )
M induced by A0 M is a quasi-isomorphism.
p,q
Proof. Write TP
= Tot(A, ). Let x Ker(d0T ) represent a cohomology class
i
. Write x =
i=n,...,0 xi with xi Ai . Assume n > 0. Then xn is in the
n
kernel of dAn and maps to zero in the cohomology of An1 (because it maps to an
element which is the boundary of xn1 up to sign). The condition on exactness
of kernels of differentials implies that the cohomology class of xn is in the image
of H n (An+1 ) H n (An ) (details omitted). Thus we can modify x by a boundary
and reach the situation where xn is a boundary. Modifying x once more we see
that we may assume xn = 0. By induction we see that every cohomology class
is represented by a cocycle x = x0 . Finally, the condition on exactness of kernels
tells us two such cocycles x0 and x00 are cohomologous if and only if their image in
H 0 (M ) are the same.
Lemma 22.11. Let M be a complex of abelian groups. Let

0 M A0 A1 A2 . . .
be an exact complex of complexes of abelian groups such that for all p Z the
complexes
0 Coker(dpM ) Coker(dpA ) Coker(dpA ) Coker(dpA ) . . .
0
to obtain a double complex. Let Tot (A, ) be the

are exact as well. Set A =
product total complex associated to the double complex (see proof ). Then the map
M Tot (A, ) induced by M A0 is a quasi-isomorphism.
p,q
Aqp
Proof. Abbreviating T = Tot (A, ) we define

Y
Y
Tn =
Ap,q =
p+q=n
p+q=n
Aqp
As differential we use
d((xp,q )) = (fp (xp1,q ) + (1)p dAp (xp,q1 ))
Let x Ker(d0T ) represent a cohomology class H 0 (T ). Write x = (xi ) with
1
xi Ai
A01 ). Hence we see that
i . Note that x0 maps to zero in Coker(A1
x0 = m0 + d(y) for some m0 M 0 . Then d(m0 ) = 0 because d(x0 ) = 0 as x is
a cocycle. Thus, replacing by something in the image of H 0 (M ) H 0 (T ) we
0
may assume that x0 is in the image of d : A1
0 A0 .
0
Assume x0 Im(A1
0 A0 ). We claim that in this case = 0. To prove this we
find, by induction on n elements y1 , . . . , yn with yi Ai1
such that x0 = d(y0 )
i
and xj = fj1 (yj1 ) + (1)j d(yj ). This is clear for n = 0. Proof of induction step
is omitted. Taking y = (yi ) we find that d(y) = .
This shows that H 0 (M ) H 0 (T ) is surjective. We omit the proof of injectivity.

HOMOLOGICAL ALGEBRA
55
23. Injectives
Definition 23.1. Let A be an abelian category. An object J Ob(A) is called
injective if for every injection A , B and every morphism A J there exists a
morphism B J making the following diagram commute
/B
A

J
Here is the obligatory characterization of injective objects.
Lemma 23.2. Let A be an abelian category. Let I be an object of A. The following
are equivalent:
(1) The object I is injective.
(2) The functor B 7 HomA (B, I) is exact.
(3) Any short exact sequence
0IAB0
in A is split.
(4) We have ExtA (B, I) = 0 for all B Ob(A).
Proof. Omitted.
Lemma 23.3. Let

Q A be an abelian category. Suppose I , is a set of injective
objects of A. If I exists then it is injective.
Proof. Omitted.
Definition 23.4. Let A be an abelian category. We say A has enough injectives

if every object A has an injective morphism A J into an injective object J.
Definition 23.5. Let A be an abelian category. We say that A has functorial
injective embeddings if there exists a functor
J : A Arrows(A)
such that
(1) s J = idA ,
(2) for any object A Ob(A) the morphism J(A) is injective, and
(3) for any object A Ob(A) the object t(J(A)) is an injective object of A.
We will denote such a functor by A 7 (A J(A)).
24. Projectives
Definition 24.1. Let A be an abelian category. An object P Ob(A) is called
projective if for every surjection A B and every morphism P B there exists a
morphism P A making the following diagram commute
/B
AO
?
P
Here is the obligatory characterization of projective objects.
56
HOMOLOGICAL ALGEBRA
Lemma 24.2. Let A be an abelian category. Let P be an object of A. The following

are equivalent:
(1) The object P is projective.
(2) The functor B 7 HomA (P, B) is exact.
(3) Any short exact sequence
0ABP 0
in A is split.
(4) We have ExtA (P, A) = 0 for all A Ob(A).
Proof. Omitted.
Lemma 24.3. Let A be `

an abelian category. Suppose P , is a set of
projective objects of A. If P exists then it is projective.
Proof. Omitted.
Definition 24.4. Let A be an abelian category. We say A has enough projectives

if every object A has an surjective morphism P A from an projective object P
onto it.
Definition 24.5. Let A be an abelian category. We say that A has functorial
projective surjections if there exists a functor
P : A Arrows(A)
such that
(1) t J = idA ,
(2) for any object A Ob(A) the morphism P (A) is surjective, and
(3) for any object A Ob(A) the object s(P (A)) is an projective object of A.
We will denote such a functor by A 7 (P (A) A).
25. Injectives and adjoint functors
Here are some lemmas on adjoint functors and their relationship with injectives.
See also Lemma 7.3.
Lemma 25.1. Let A and B be abelian categories. Let u : A B and v : B A
be additive functors. Assume
(1) u is right adjoint to v, and
(2) v transforms injective maps into injective maps.
Then u transforms injectives into injectives.
Proof. Let I be an injective object of A. Let : N M be an injective map in B
and let : N uI be a morphism. By adjointness we get a morphism : vN I
and by assumption v : vN vM is injective. Hence as I is an injective object we
get a morphism : vM I extending . By adjointness again this corresponds to
a morphism : M uI as desired.

Remark 25.2. Let A, B, u : A B and v : B A be as in Lemma 25.1. In
the presence of assumption (1) assumption (2) is equivalent to requiring that v is
exact. Moreover, condition (2) is necessary. Here is an example. Let A B be a
ring map. Let u : ModB ModA be u(N ) = NA and let v : ModA ModB be
v(M ) = M A B. Then u is right adjoint to v, and u is exact and v is right exact,
HOMOLOGICAL ALGEBRA
57
but v does not transform injective maps into injective maps in general (i.e., v is
not left exact). Moreover, it is not the case that u transforms injective B-modules
into injective A-modules. For example, if A = Z and B = Z/pZ, then the injective
B-module Z/pZ is not an injective Z-module. In fact, the lemma applies to this
example if and only if the ring map A B is flat.
(1)
(2)
(3)
(4)
u is right adjoint to v,
v transforms injective maps into injective maps,
A has enough injectives, and
vB = 0 implies B = 0 for any B Ob(B).
Then B has enough injectives.

Proof. Pick B Ob(B). Pick an injection vB I for I an injective object of A.
According to Lemma 25.1 and the assumptions the corresponding map B uI is
the injection of B into an injective object.

Remark 25.4. Let A, B, u : A B and v : B A be as In Lemma 25.3. In the
presence of conditions (1) and (2) condition (4) is equivalent to v being faithful.
Moreover, condition (4) is needed. An example is to consider the case where the
functors u and v are both the zero functor.
(1)
(2)
(3)
(4)
(5)
u is right adjoint to v,
v transforms injective maps into injective maps,
A has enough injectives,
vB = 0 implies B = 0 for any B Ob(B), and
A has functorial injective hulls.
Then B has functorial injective hulls.

Proof. Let A 7 (A J(A)) be a functorial injective hull on A. Then B 7 (B
uJ(vB)) is a functorial injective hull on B. Compare with the proof of Lemma
25.3.

Lemma 25.6. Let A and B be abelian categories. Let u : A B be a functor. If
there exists a subset P Ob(B) such that
(1) every object of B is a quotient of an element of P, and
(2) for every P P there exists an object Q of A such that HomA (Q, A) =
HomB (P, u(A)) functorially in A,
then there exists a left adjoint v of u.
Proof. By the Yoneda lemma (Categories, Lemma 3.5) the object Q of A corresponding to P is defined up to unique isomorphism by the formula HomA (Q, A) =
HomB (P, u(A)). Let us write Q = v(P ). Denote iP : P u(v(P )) the map corresponding to idv(P ) in HomA (v(P ), v(P )). Functoriality in (2) implies that the
bijection is given by
HomA (v(P ), A) HomB (P, u(A)),
7 u() iP
58
HOMOLOGICAL ALGEBRA
For any pair of elements P1 , P2 P there is a canonical map

HomB (P2 , P1 ) HomA (v(P2 ), v(P1 )),
7 v()
which is characterized by the rule u(v()) iP2 = iP1 in HomB (P2 , u(v(P1 ))).
Note that 7 v() is compatible with composition; this can be seen directly from
the characterization. Hence P 7 v(P ) is a functor from the full subcategory of B
whose objects are the elements of P.
Given an arbitrary object B of B choose an exact sequence
P2 P 1 B 0
which is possible by assumption (1). Define v(B) to be the object of A fitting into
the exact sequence
v(P2 ) v(P1 ) v(B) 0
Then
HomA (v(B), A) = Ker(HomA (v(P1 ), A) HomA (v(P2 ), A))
= Ker(HomB (P1 , u(A)) HomB (P2 , u(A)))
= HomB (B, u(A))
Hence we see that we may take P = Ob(B), i.e., we see that v is everywhere
defined.

26. Essentially constant systems
In this section we discuss essentially constant systems with values in additive categories.
Lemma 26.1. Let I be a category, let A be a pre-additive Karoubian category, and
let M : I A be a diagram.
(1) Assume I is filtered. The following are equivalent
(a) M is essentially constant,
(b) X = colim M exists and there exists a cofinal filtered subcategory I 0
I and for i0 Ob(I 0 ) a direct sum decomposition Mi0 = Xi0 Zi0 such
that Xi0 maps isomorphically to X and Zi0 to zero in Mi00 for some
i0 i00 in I 0 .
(2) Assume I is cofiltered. The following are equivalent
(a) M is essentially constant,
(b) X = lim M exists and there exists an initial cofiltered subcategory I 0
I and for i0 Ob(I 0 ) a direct sum decomposition Mi0 = Xi0 Zi0 such
that X maps isomorphically to Xi0 and Mi00 Zi0 is zero for some
i00 i0 in I 0 .
Proof. Assume (1)(a), i.e., I is filtered and M is essentially constant. Let X =
colim Mi . Choose i and X Mi as in Categories, Definition 22.1. Let I 0 be the full
subcategory consisting of objects which are the target of a morphism with source
i. Suppose i0 Ob(I 0 ) and choose a morphism i i0 . Then X Mi Mi0
composed with Mi0 X is the identity on X. As A is Karoubian, we find a direct
summand decomposition Mi0 = Xi0 Zi0 , where Zi0 = Ker(Mi0 X) and Xi0
maps isomorphically to X. Pick i k and i0 k such that Mi0 X Mi Mk
equals Mi0 Mk as in Categories, Definition 22.1. Then we see that Mi0 Mk
annihilates Zi0 . Thus (1)(b) holds.
HOMOLOGICAL ALGEBRA
59
Assume (1)(b), i.e., I is filtered and we have I 0 I and for i0 Ob(I 0 ) a direct sum
decomposition Mi0 = Xi0 Zi0 as stated in the lemma. To see that M is essentially
constant we can replace I by I 0 , see Categories, Lemmas 22.8 and 17.2. Pick any
i Ob(I) and denote X Mi the inverse of the isomorphism Xi X followed
by the inclusion map Xi Mi . If j is a second object, then choose j k such
that Zj Mk is zero. Since I is filtered we may also assume there is a morphism
i k (after possibly increasing k). Then Mj X Mi Mk and Mj Mk
both annihilate Zj . Thus after postcomposing by a morphism Mk Ml which
annihilates the summand Zk , we find that Mj X Mi Ml and Mj Ml
are equal, i.e., M is essentially constant.
The proof of (2) is dual.
Lemma 26.2. Let I be a category. Let A be an additive, Karoubian category. Let

F : I A and G : I A be functors. The following are equivalent
(1) colimI F G exists, and
(2) colimI F and colimI G exist.
In this case colimI F G = colimI F colimI G.
Proof. Assume (1) holds. Set W = colimI F G. Note that the projection onto
F defines natural tranformation F G F G which is idempotent. Hence we
obtain an idempotent endomorphism W W by Categories, Lemma 14.7. Since
A is Karoubian we get a corresponding direct sum decomposition W = X Y , see
Lemma 4.2. A straightforward argument (omitted) shows that X = colimI F and
Y = colimI G. Thus (2) holds. We omit the proof that (2) implies (1).

Lemma 26.3. Let I be a filtered category. Let A be an additive, Karoubian category. Let F : I A and G : I A be functors. The following are equivalent
(1) F G : I A is essentially constant, and
(2) F and G are essentially constant.
Proof. Assume (1) holds. In particular W = colimI F G exists and hence by
Lemma 26.2 we have W = X Y with X = colimI F and Y = colimI G. A
straightforward argument (omitted) using for example the characterization of Categories, Lemma 22.6 shows that F is essentially constant with value X and G is
essentially constant with value Y . Thus (2) holds. The proof that (2) implies (1)
is omitted.

27. Inverse systems
Let C be a category. In Categories, Section 21 we defined the notion of an inverse
system over a partially ordered set (with values in the category C). If the partially
ordered set is N = {1, 2, 3, . . .} with the usual ordering such an inverse system
over N is often simply called an inverse system. It consists quite simply of a pair
(Mi , fii0 ) where each Mi , i N is an object of C, and for each i > i0 , i, i0 N
a morphism fii0 : Mi Mi0 such that moreover fi0 i00 fii0 = fii00 whenever this
makes sense. It is clear that in fact it suffices to give the morphisms M2 M1 ,
M3 M2 , and so on. Hence an inverse system is frequently pictured as follows
2
M1 M2 M3 . . .
Moreover, we often omit the transition maps i from the notation and we simply
say let (Mi ) be an inverse system.
60
HOMOLOGICAL ALGEBRA
The collection of all inverse systems with values in C forms a category with the
obvious notion of morphism.
Lemma 27.1. Let C be a category.
(1) If C is an additive category, then the category of inverse systems with values
in C is an additive cateogry.
(2) If C is an abelian category, then the category of inverse systems with values
in C is an abelian cateogry. A sequence (Ki ) (Li ) (Mi ) of inverse
systems is exact if and only if each Ki Li Ni is exact.
Proof. Omitted.
The limit (see Categories, Section 21) of such an inverse system is denoted lim Mi ,
or limi Mi . If C is the category of abelian groups (or sets), then the limit always
exists and in fact can be described as follows
Y
limi Mi = {(xi )
Mi | i (xi ) = xi1 , i = 2, 3, . . .}
see Categories, Section 15. However, given a short exact sequence
0 (Ai ) (Bi ) (Ci ) 0
of inverse systems of abelian groups it is not always the case that the associated
system of limits is exact. In order to discuss this further we introduce the following
notion.
Definition 27.2. Let C be an abelian category. We say the inverse system (Ai )
satisfies the Mittag-Leffler condition, or for short is ML, if for every i there exists
a c = c(i) i such that
Im(Ak Ai ) = Im(Ac Ai )
for all k c.
It turns out that the Mittag-Leffler condition is good enough to ensure that the
lim-functor is exact, provided one works within the abelian category of abelian
groups, or abelian sheaves, etc. It is shown in a paper by A. Neeman (see [Nee02])
that this condition is not strong enough in a general abelian category (where limits
of inverse systems exist).
Lemma 27.3. Let
0 (Ai ) (Bi ) (Ci ) 0
be a short exact sequence of inverse systems of abelian groups.
(1) In any case the sequence
0 limi Ai limi Bi limi Ci
is exact.
(2) If (Bi ) is ML, then also (Ci ) is ML.
(3) If (Ai ) is ML, then
0 limi Ai limi Bi limi Ci 0
is exact.
Proof. Nice exercise. See Algebra, Lemma 84.1 for part (3).
HOMOLOGICAL ALGEBRA
61
Lemma 27.4. Let

(Ai ) (Bi ) (Ci ) (Di )
be an exact sequence of inverse systems of abelian groups. If the system (Ai ) is ML,
then the sequence
limi Bi limi Ci limi Di
is exact.
Proof. Let Zi = Ker(Ci Di ) and Ii = Im(Ai Bi ). Then lim Zi = Ker(lim Ci
lim Di ) and we get a short exact sequence of systems
0 (Ii ) (Bi ) (Zi ) 0
Moreover, by Lemma 27.3 we see that (Ii ) has (ML), thus another application of
Lemma 27.3 shows that lim Bi lim Zi is surjective which proves the lemma.
The following characterization of essentially constant inverse systems shows in particular that they have ML.
Lemma 27.5. Let A be an abelian category. Let (Ai ) be an inverse system in A
with limit A = lim Ai . Then (Ai ) is essentially constant (see Categories, Definition
22.1) if and only if there exists an i and for all j i a direct sum decomposition
Aj = A Zj such that (a) the maps Aj 0 Aj are compatible with the direct sum
decompositions, (b) for all j there exists some j 0 j such that Zj 0 Zj is zero.
Proof. Assume (Ai ) is essentially constant. Then there exists an i and a morphism
Ai A such that for all j i there exists a j 0 j such that Aj 0 Aj factors
as Aj 0 Ai A Aj (the last map comes from A = lim Ai ). Hence setting
Zj = Ker(Aj A) for all j i works. Proof of the converse is omitted.

Lemma 27.6. Let
0 (Ai ) (Bi ) (Ci ) 0
be an exact sequence of inverse systems of abelian groups. If (Ai ) has ML and (Ci )
is essentially constant, then (Bi ) has ML.
Proof. After renumbering we may assume that Ci = C Zi compatible with
transition maps and that for all i there exists an i0 i such that Zi0 Zi is zero, see
Lemma 27.5. Pick i. Let c i by an integer such that Im(Ac A) = Im(Ai0 Ai )
for all i0 c. Let c0 c be an integer such that Zc0 Zc is zero. For i0 c0
consider the maps
/ Bi0
/ C Z i0
/0
/ Ai 0
0

/ Ac0

/ Bc0

/ C Zc 0
/0

/ Ac

/ Bc

/ C Zc
/0

/ Ai

/ Bi

/ C Zi
/0
Because Zc0 Zc is zero the image Im(Bc0 Bc ) is an extension C by a subgroup

A0 Ac which contains the image of Ac0 Ac . Hence Im(Bc0 Bi ) is an
62
HOMOLOGICAL ALGEBRA
extension of C by the image of A0 which is the image of Ac Ai by our choice of

c. In exactly the same way one shows that Im(Bi0 Bi ) is an extension of C by
the image of Ac Ai . Hence Im(Bc0 Bi ) = Im(Bi0 Bi ) and we win.

The correct version of the following lemma is More on Algebra, Lemma 60.2.
Lemma 27.7. Let
(A2
A1
A0i A1i )
i
i
be an inverse system of complexes of abelian groups and denote A2 A1
A0 A1 its limit. Denote (Hi1 ), (Hi0 ) the inverse systems of cohomologies, and
1
denote H 1 , H 0 the cohomologies of A2 A1 A0 A1 . If (A2
i ) and (Ai )
1
0
0
are ML and (Hi ) is essentially constant, then H = lim Hi .
) and Iij = Im(Aij1 Aji ). Note that lim Zi0 =
Proof. Let Zij = Ker(Aji Aj+1
i
0
1
Ker(lim Ai lim Ai ) as taking kernels commutes with limits. The systems (Ii1 )
1
and (Ii0 ) have ML as quotients of the systems (A2
i ) and (Ai ), see Lemma 27.3.
Thus an exact sequence
0 (Ii1 ) (Zi1 ) (Hi1 ) 0
of inverse systems where (Ii1 ) has ML and where (Hi1 ) is essentially constant by
assumption. Hence (Zi1 ) has ML by Lemma 27.6. The exact sequence
0
0 (Zi1 ) (A1
i ) (Ii ) 0
and an application of Lemma 27.3 shows that lim A1

lim Ii0 is surjective. Fii
nally, the exact sequence
0 (Ii0 ) (Zi0 ) (Hi0 ) 0
and Lemma 27.3 show that lim Ii0 lim Zi0 lim Hi0 0 is exact. Putting
everything together we win.

Sometimes we need a version of the lemma above where we take limits over big
ordinals.
Lemma 27.8. Let be an ordinal. Let K , < be an inverse system of
complexes of abelian groups over . If for all < the complex K is acyclic and
the map
Kn lim< Kn
is surjective, then the complex lim< K is acyclic.
Proof. By transfinite induction we prove this holds for every ordinal and every
system as in the lemma. In particular, whilst proving the result for we may
assume the complexes lim< Kn are acyclic.
Let x lim< K0 with d(x) = 0. We will find a y K1 with d(y) = x. Write
x = (x ) where x K0 is the image of x for < . We will construct y = (y )
by transfinite induction.
For = 0 let y0 K01 be any element with d(y0 ) = x0 .
For = + 1 a successor, we have to find an element y which maps both to
y by the transition map f : K K and to x under the differential. As a
first approximation we choose y0 with d(y0 ) = x . Then the difference y f (y0 )
is in the kernel of the differential, hence equal to d(z ) for some z K2 . By
HOMOLOGICAL ALGEBRA
63
assumption, the map f 2 : K2 K2 is surjective. Hence we write z = f (z )

and change y0 into y = y0 + d(z ) which works.
If is a limit ordinal, then we have the element (y )< in lim< K1 whose
differential is the image of x . Thus we can argue in exactly the same manner as
above using the termwise surjective map of complexes f : K lim< K and
the fact (see first paragraph of proof) that we may assume lim< K is acyclic by
induction.

28. Exactness of products
Lemma 28.1. Let I be a set. For i I let Li Mi Ni be a complex of abelian
groups. Let Hi = Ker(Mi Ni )/Im(Li Mi ) be the cohomology. Then
Y
Y
Y
Li
Mi
Ni
Q
is a complex of abelian groups with homology Hi .
Proof. Omitted.

29. Other chapters
Preliminaries
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)
(21)
(22)
(23)
(24)
Introduction
Conventions
Set Theory
Categories
Topology
Sheaves on Spaces
Sites and Sheaves
Stacks
Fields
Commutative Algebra
Brauer Groups
Homological Algebra
Derived Categories
Simplicial Methods
More on Algebra
Smoothing Ring Maps
Sheaves of Modules
Modules on Sites
Injectives
Cohomology of Sheaves
Cohomology on Sites
Differential Graded Algebra
Divided Power Algebra
Hypercoverings
Schemes
(25) Schemes
(26) Constructions of Schemes
(27) Properties of Schemes
(28)
(29)
(30)
(31)
(32)
(33)
(34)
(35)
(36)
(37)
(38)
(39)
(40)
Morphisms of Schemes
Cohomology of Schemes
Divisors
Limits of Schemes
Varieties
Topologies on Schemes
Descent
Derived Categories of Schemes
More on Morphisms
More on Flatness
Groupoid Schemes
More on Groupoid Schemes
Etale
Morphisms of Schemes
Topics in Scheme Theory

(41)
(42)
(43)
(44)
(45)
(46)
Chow Homology
Adequate Modules
Dualizing Complexes
Etale
Cohomology
Crystalline Cohomology
Pro-etale Cohomology
Algebraic Spaces
(47)
(48)
(49)
(50)
(51)
(52)
(53)
Algebraic Spaces
Properties of Algebraic Spaces
Morphisms of Algebraic Spaces
Decent Algebraic Spaces
Cohomology of Algebraic Spaces
Limits of Algebraic Spaces
Divisors on Algebraic Spaces
64
HOMOLOGICAL ALGEBRA
(54)
(55)
(56)
(57)
(58)
(59)
(60)
(61)
(62)
Algebraic Spaces over Fields

Topologies on Algebraic Spaces
Descent and Algebraic Spaces
Derived Categories of Spaces
More on Morphisms of Spaces
Pushouts of Algebraic Spaces
Groupoids in Algebraic Spaces
More on Groupoids in Spaces
Bootstrap
Topics in Geometry
(63)
(64)
(65)
(66)
(67)
Quotients of Groupoids
Simplicial Spaces
Formal Algebraic Spaces
Restricted Power Series
Resolution of Surfaces
Deformation Theory
(68) Formal Deformation Theory
(69) Deformation Theory
(70) The Cotangent Complex
Algebraic Stacks
(71) Algebraic Stacks

(72) Examples of Stacks
(73) Sheaves on Algebraic Stacks
(74) Criteria for Representability
(75) Artins Axioms
(76) Quot and Hilbert Spaces
(77) Properties of Algebraic Stacks
(78) Morphisms of Algebraic Stacks
(79) Cohomology of Algebraic Stacks
(80) Derived Categories of Stacks
(81) Introducing Algebraic Stacks
Miscellany
(82) Examples
(83) Exercises
(84) Guide to Literature
(85) Desirables
(86) Coding Style
(87) Obsolete
(88) GNU Free Documentation License
(89) Auto Generated Index
References
Pierre Deligne, Th
eorie de Hodge. II, Inst. Hautes Etudes
Sci. Publ. Math. (1971), no. 40,
557.
[Eis95] David Eisenbud, Commutative algebra, Graduate Texts in Mathematics, vol. 150,
Springer-Verlag, 1995.
[Lan02] Serge Lang, Algebra, third ed., Graduate Texts in Mathematics, vol. 211, SpringerVerlag, New York, 2002.
[McC01] John McCleary, A users guide to spectral sequences, second ed., Cambridge Studies in
Advanced Mathematics, vol. 58, Cambridge University Press, Cambridge, 2001.
[ML63] Saunders Mac Lane, Homology, Die Grundlehren der mathematischen Wissenschaften,
Bd. 114, Academic Press Inc., Publishers, New York, 1963.
[Nee02] Amnon Neeman, A counterexample to a 1961 theorem in homological algebra, Invent.
Math. 148 (2002), no. 2, 397420, With an appendix by P. Deligne.
[Ser53] Jean-Pierre Serre, Groupes dhomotopie et classes de groupes ab
eliens, Ann. of Math.
(2) 58 (1953), 258294.
[Del71]

This Is A Chapter of The Stacks Project, Version 4287507, Compiled On Oct 20, 2014

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

This Is A Chapter of The Stacks Project, Version 4287507, Compiled On Oct 20, 2014

Uploaded by

Copyright:

Available Formats

HOMOLOGICAL ALGEBRA

Definition 3.3. In a preadditive category A we call zero object, and we denote it

where the diagonal compositions are zero. It follows that i p + j q : z z is

Then we can form the map a p + b q : z w. In this `

Lemma 3.10. Let C be a preadditive category. Let xy with morphisms i, j, p, q as

isomorphism x = y z as in Remark 3.6. Thus (1) (3). The implication (2)

Definition 5.3. Let f : x y be a morphism in an abelian category.

In an abelian category, if x y is a subobject, then we denote

Definition 5.9. Let A be an abelian category. Let i : A B and q : B C be

(2) Given a morphism : B A right inverse to A B, there exists a unique

Lemma 5.11. Let A be an abelian category. Let

Lemma 5.13. Let A be an abelian category. Let

Lemma 5.15. Let A be an abelian category. Let f : x y and g : y z be

(1) The sequence x y z is exact.

be a commutative diagram with exact rows.

Ker() Ker() Ker() Coker() Coker() Coker()

diagram with exact rows

As l 0 = i = 0 and as the third row of the diagram above is exact,

sequence p Ker() 0 and d yields an object s, an epimorphism m : s r and

Lemma 5.15 to the exact sequence x u Coker() and a n yields an object

be a commutative diagram with exact rows. Then, the induced diagram

Lemma 5.19. Let A be an abelian category. Let

be a commutative diagram with exact rows.

to conclude that Ker(y z) Ker(y 0 z 0 ) is surjective. Finally, we apply

be a commutative diagram with exact rows. If , are isomorphisms,  is injective,

(A, B) 7 ExtA (A, B)

The extension E 0 is called the pullback of E via B 0 B.

Note that if E1 and E2 are extensions of B by A, then E1 E2 is an extension of

The extension E1 + E2 is called the Baer sum of the given extensions.

Lemma 6.4. Let A be an abelian category. Let 0 M1 M2 M3 0 be a

for all objects N of A, and

for all objects N of A.

F (i)F (a)+F (j)F (b)

Lemma 7.2. Let A and B be abelian categories. Let F : A B be an exact

which is equivalent to (s0 )1 (f 0 + g 0 ).

s : Y 0 Y 00 in S and morphisms a00 , b00 : X 0 Y 00 such that the following diagram

 a00 ,b00 ,a00 +b00 

Proof. Assume S is a left multiplicative system. Let a : X Y be a morphism

Moreover, a Serre subcategory is an abelian category and the inclusion functor is

Lemma 9.3. Let A be an abelian category. Let C be a subcategory of A. Then C

Lemma 9.4. Let A, B be abelian categories. Let F : A B be an exact functor.

Definition 9.5. Let A, B be abelian categories. Let F : A B be an exact

with t S. Set W = C qA B = Coker((t, g) : A C B). Then Ker(t) Ker(s)

Finally, consider morphisms f, g : B C and a morphism s : A B in S

with [A B C] 7 [B] [A] [C] of K0 (A). The short exact sequence

Suppose we are given an object M of an abelian category A and a complex of the

In this situation we define

and H 1 (M, , ) = Ker()/Im().

Lemma 10.3. Let A be an abelian category. Let C A be a Serre subcategory and

([Bi ] [Ai ] [Ci ])

([Bi ] [Ai ] [Ci ])

in the free abelian group on the objects of A. We can rewrite this as

By using we see that it is equal to the object

Consider the map

(1) for every short exact sequence as above the sequence

the short exact sequence 0 A B C 0 and the -functor F gives that

/ Coker(Gn (B) Gn (C))

commutes. This is clearly uniquely determined by the requirements imposed. We

Lemma 12.1. Let A be an additive category. Let f, g : B C be morphisms

(1) If the maps f, g : A B are homotopic, then the induced maps Hi (f )

Lemma 12.6. Let A be an abelian category. Suppose that

A cochain complex A in an additive category A is a complex

a pair of morphisms of cochain complexes f, g : A B is is a collection of

be a commutative diagram with exact rows. If , are isomorphisms, is injective,

a00 ,b00 ,a00 +b00